-------
Introduction

  A significant  fraction of the  aerosols in the  atmosphere is
carbonaceous.   Particulate carbon may be  classified  into three
components:  organic  carbon (OC) ,  elemental carbon   (EC), and
carbonate  or  inorganic carbon  (CC) .  Particulate OC  and  EC are
emitted directly from  combustion  sources.  Organic aerosols may
also   be   formed   by  photochemical   oxidation  of  gaseous
hydrocarbons  in the  atmosphere1.  Carbonates  origninate  froi
soil,  though  some  found  on quartz or  glass fiber filters may
result  from  reaction  of  CO2 with  filter substrate  alkalinity
during sampling.

  Particulate  carbon have  been  investigated for  its potential
effects  on  health,   visibility,   climate,   and  catalysis  of
atmospheric  oxidative  processes.  Organics   in  urban  aerosols
have been  shown  to be mutagenic2. Soot  particles  caused cancer
in laboratory  animals-*.  The presence  of soot in air  is a major
cause  of  visibility  reduction  in urban areas4'5'6'7.  Recently
organic aerosols  have been  suggested  as major  contributors to
visibility  impairment  in  scenic areas  of  the  western  United
States8.  As  EC  absorbs  light,  it  may  affect  the  earth's
radiation  balance  and impact global  climate9'10.   EC has been
also investigated  for its role in catalyzing the  conversion of
atmospheric   SO2   to  sulfate11.   Thus,  the   importance  of
particulate carbon  in human health and  air  quality underscores
the need  for  the  assessment and control  of its levels  in the
atmosphere.

  Carbonaceous aerosols are  routinely collected  on quartz fiber
filters for  measurement of  OC and EC  using thermal oxidative
techniques12.  However,  the  filter sampling technique  is subject
to artifacts.   Sampling  with  tandem quartz  filters  in  Luray,
Virginia and Warren,  Michigan,  Cadle et al  found  a significant
amount  of  OC  on the back filter13. This  effect has also been
observed for samples taken during the 1987  Southern  California
Air  Quality Study  (SCAQS) .    Some back filters  had  organic
carbon  levels  greater  than  50%  of those  in the  front filter,
though  the normal  range  was  "10-35%.   These  findings  may be
interpreted as adsorption  by the quartz filter  of  gas phase OC
that was  originally present in  the particles collected  on the
front  filter,  or in  the  sampling air stream.   The  loss  of OC
from the  front filter thus would represent  a negative sampling
artifact.   On  the other hand,  a positive  artifact  would result
if the  filter  adsorbs gaseous  OC from the air stream. It could
also be possible that  both  processes occur simultaneously. Much
research is still being performed to  investigate these sampling
artifacts.

    Another source  of  positive  artifact  may  result from the use
of certain filter holder parts,  or  material  used for  packaging,
such as silicon gaskets,  rubber 0-rings, polyethylene closure
caps or bags.  The author's  laboratory had  experienced high OC

filter  blanks  after  using  filter  holders  which  had  these
materials.

  A study was  performed to  determine the nature  of the organics
collected  by  the  back  filter  and  to  identify   if  certain
material  used  in  or with  a  filter holder  could result  in OC
                              370

-------
*tif
    cts.  This paper reports the findings of this study.

  Cental Methods


                        r Wlth  a filter  nolder ca"se  an OC
     ture;  Furth*       appreciable  bleed at  normal ambient
     f UtSr  Thn    I   he bleed °f 9aseous OC must be adsorbed
       «er.  Thus,  two separate experiments were performed.
      -  x w*      5  °f  a material» the  experimental set-up
      four T^K US •    nlt:  consisted of  a   half-inch  brass
     ds.   Thr  •  I1"  J-ength'  with  compression fittings  at
      5840A L*1PPv'  located  inside the oven  of  a  Hewlett-
     loi» detectSr «-r    >,°9raph'  WaS connected  to  the  flame
     *le   or  H*  \a  a 4-P°rt  valve, which  served  to isolate
     to  the FTn   !r *i    carrier  9as sweeping through  the
             t,V i /-,/ °W  restrictor, consisting  of  a  short
              -d  1/16-inch stainless  steel  tubing,  was  used
              now  surges from  extinguishing  the  detector
        set   '® sEwltchln9-   The  He carrier and make-up  flow
            to  25 and 10 ml/min respectively.
                  r^     toluene' followed by  isopropanol
     at  20o°c    tn  tt,   PTPnriKr  t0  assembly-  The Astern  was
N if The oven w^  iihe  ?D back9r°und trace  was clean  and
^le°f the ma?erTa? f11™** to c°o1-  By uncoupling one end,  a
      and the  « H       investigation was inserted into  the
      the ninnit    WaS  resealed-  After  about  15 min  of
       v«lve Sas  ^  lso.14.at:ed  and heated  to  30°C for  one
        to the L^   ^ swltuched to  direct  the  affluent from
        the oven £     rVThe bleed at 5°°C  Was also tested
        ket  ^"temperature. The materials  tested included
        ^ina rubber 0-ring,  polyethylene  caps,  Ziplok bags,
        — Perri ,* •    °f  gas  Phase OC'  quartz filters (47 mm
      45°°,  and RR1So   '  and  Precut aluminum foil were baked
       their no   •  C resPectively  to  ensure no residual OC
        On  was  i ln  the exPeriraent.    The material  under
         then  uv       lnto  the  Petri dish  with a  quartz
         its    raPPed tightly with  aluminum foil.  The petri
         he  ou  ontents  was   baked  at   50°c  for  8   hours
    ^ -«*eri 1?Qartz  filter was removed  for  analysis of OC12!
    e *  n°t in  p"®l?rej.th?t only  aliquots  from the portion of
        alysis   T    W1-th the  susPected material  was taken
         a PolvpA*r,?  addltion  to  the   materials  mentioned
             ycarbonate  filter holder was  also  tested.   In

         9asketrt2n  ff1ter  was  assembled   into  the  holder
         f°il  a/^  °~rin9/  or  caP-    The  entire  unit  was
               na  tested as described.

               en  at- r^ong  Beach,  CA,  was  analyzed using  a
        -.---"•iuot ~f 4-v. GC.~MS   to   determine  the  organics
       r, 5°°C  anri      e fllter was thermally desorbed in He
       f!lu«m.  Tu  cryo~focussed at the head  of  a 40m DB-1
          23QOC  ®  oven  temperature  was  ramped  linearily
           of  th    15 /min  to  effect the  separation.   The
           Ubrarv   JomP°unds  was based  on  matching mass
               «*y  of  45,000 compounds.
                          371

-------
Results and Discussion
                                                       30° &t
  The organic vapor  emission  from a silicon gasket at J   $
50°C is  shown  in Figure  2.  The spike near  time zero ' *   ^
result of valve switching. The broad peak that  followed w ^0$,
gas phase organic vapor accumulated  inside the  nipple at   ^
The concentration declined gradually,  and finally reacn   tfas
steady state  level  in about  20 min. When the  temperatu*  ^
raised to 50°C,  the emission level rose almost  immediate **yp
dip  which  followed  was  probably   due  to   some   tr  ^$
temperature  fluctuation   of  the oven  and the  nipple.    ut '
                                                      .
&
appro               ,                        ,
times  higher  than at  30°C,  was  reached.  When the nipt^
isolated  by switching  of  the valve,  the level  retuj"^s,
slowly  but eventually  to  the  starting  baseline.      to
organic  vapor  emitted  by  silicon  tended  to  cling
surfaces   of   transfer   lines   and  valve.   In  contrae '
emission  from Viton  was very low,  and also  "nonsticwr
signal  returned quickly to  the  baseline  when the sou
isolated.   By  comparing  steady  state  levels,  the    t
organic  emission  of  the substances tested at  50  C wer  ^ >
as follows: Viton, 1,  Polyethylene cap, 2, Polyethylene  ^
rubber  0-ring, 8,  and  silicon   gasket,  10.  In   othe
silicon  outgased  ten  times as much as  Viton  under tne
experimental  conditions.  If contamination from  silic° f
is  suspected,  substituting the  gasket  with  one made
should significantly reduce the problem.
                                                      f
  The  results  of  total  organic  carbon analysis  o
filters  that  had been  packaged  with a  filter holder ^ ^ .
are  given in  Table I.   It is  clear that large amount
phase  OC was  emitted  from polymeric materials, such a
or rubber,  and was adsorbed by the  filter. Both the Pf
bag  were made  of  Polyethylene and the amount  of _ TO  ^
both  filters  was  similar.     It  was   a  surprise    d .9
Polycarbonate   filter  holder   body  also   had
contamination  of  the  filter.  Perhaps  the holder
totally  clean and  still contained some  residual
from the silicon gaskets or  rubber  0-rings used previ
the  holder •
   The GC-MS  analysis of  back quartz  filters
 number of  compounds,  many  of  which  were  oxygen,
 and/or phosphorus-containing.  The major ones  inclua
 phosphate,   diethylphthalate,  butylcyclohexyiP
 diethyloxalate,  1,2,3-propanetriol  triacetate,
 trimethylpentyl-2-methylpropionate,   methylhexa«-
 methylethyltetradecanoate,  benzaldehyde,
 pyrrolidinone,   2,6-bis(1,1-dimethylethyl)-4-m
 acetic acid,  isoquinoline  and 3,5-dimethylpyridine

   Recently,  using a denuder difference  technique  whi^t2 *J
 a parallel plate denuder  of  quartz filter strips,  pfe$e<
 that the  adsorption of organics  during  sampling   und*\
 positive artifact14.  The  polar nature of the comp   t&0
 in the back  filter under  the present  study  sugge>sof *$*
 positive sampling  artifact is perhaps the result  _.lfld9
 adsorption   of  certain   gas  phase  organic   cc
 interaction with the surface  hydroxyl  groups of the

                               372

-------
 Ckn°wledgement

  Th
         hor  thanks  Mr-  John  Wood  for performing  the  GC-MS
          Part  °f this research was  funded by Electric  Power
        Institute.
3
   S T   u •
   (is!???6   6r and S*K'  Friedlander' Atmos. Environ. H:  157

2,  j
    •N.  Pitts,  Environ.  Health Perspect 47: 115  (1983).


     RC  Washington Group,  Cancer Research 40: 1  (1980).

"**  W i>
   (^g*  pj-ers°n and P.A.  Russell,  Atmos. Environ. 13; 1623

5.  p
   Er>v,:  Groblic*i,  G.T. Wolff,  and R.J.  Countess, Atmos.
   ^"^AiaiU  15;  2473  (1981) .
6.

   &tUAoC°nklin'  GtR-  Cass'  L~c chu'  and E>s-  Macias, in
               Aerosol -  Source/Air Quality Relationships.
       . Marcias  and  P.K.  Hopke ed.)  Amer.  Chem. Soc.,
7       Lngton,  D.C.  (1981).
 •  g

   JAPnAatsinis'  T' Novakov,  E.G.  Ellis,  and S.K. Friedlander,
a   ^--i=a 34;  643  (1984) .


   on iaM?cias  and W-H- White,  Third  International Conference
   Oct  f"bor»aceous Particles  in the Atmosphere,  Berkeley, CA,
9     ' 5~8,  1987, Paper No.  E9.

    '  amamoto  and M. Tanaka, J. Atmos.  Sci.  29.: 1405 (1972).
   H
   ^kra^h61'  in Man/s Impact  on  Climate (W.  Bach,  J.
   &p. 220;.  and Wt Kell°gg ed-) Elsevier,  Amsterdam. 1979,

*l. *
            '  S-Gt Chang, and A.B. Hanker,  Science 186:  259



         eller/ K'Kt Fun9' S-L. Heisler,  D.  Grosjean,  and G.
         '  Atmospheric Particulate Carbon Observations in
        .    Rural Areas of the United States  in:  Particulata
               pheric Life cYcler G.T. Wolff and  R.L.
               Plenum Publishing Corp., New  York,  N.Y.,  1982
              P>J- Groblicki, and P. A. Mulawa, Atmos.
           12:  593 (1983).
           T                                   i
         i  Investigation of Particulate Organic Carbon
         i?n Artifacts by Using a Quartz Fiber Denuder
        at   . 1986 Carbonaceous Species Method Comparison
           Cltrus College, Glendora, California, Draft
       Alt-  r°Ject 1630-11 Electric Power Research Institute
         ca'  CA,  June 1987.

                           373

-------
Table I :  Total Organic Carbon Loadings on Packaged Filters
          Heated to 50°C
       Packaging	          Carbon, ug/filter

    Glass Container                   10.8

     Silicon Gasket                   37.4

     Rubber 0-ring                    31.7

       P.E. Bag                       16.2

       P.E. Cap                       17.5

     Filter Holder                    20.9
                             374

-------
                              HCLIUM CARRIER
CO
«v]
CJ1
                          HELIUM
                        MAKE-UP
>  FLOW
  RE8TRICTOR
                                                                  0-RINQ/QA8KET
                           VALVE  POSITION

                           •— EQUILIBRATION
                           	 INJECTION
                                     DETECTOR
                                Figure 1.  Apparatus to  Determine Organic Vapor  Emission

-------
                        in
                        CO
                        z
                        o
                        0-
                        CO
                        111
                        oc

                        o
                        uZ
CO
vj
O5
                                                                                Gasket isolated
                                30° C
50  C
I
0
1 1
20
1 	 1 	
40
	 1 	 1
60
Time, mln
                                            rig.ure 2.  Or&anic Vapor Emission  from a Silicon Gasket

-------
 r^OSPHERXC FATE OF VOC'S
 L« THE 1987-1988 DENVER IEMP
^HCtSVe5umSntegrated  Environmental   Management  Project   (IEMP)   is

l^lth  S^ts  iS  the  D      M ?  °       M    Potential toxic/hazardous
"Cfii ? Or +•    *-""uLri  Mf^ ^CEt" i tnss t" cfciH  f^iI*-              i • * i i   • ^***, ^*)   dEjowOXt*L GC]
ij  UQ |1   t O3f '1^*  "      ^—fc* t. j. ilia, t, cU * v>Uir6riLi V  V6rV lliT'lf* 7C! If Ti/^t»m nK\^*ii^  4*1-1

^3 "P^* Ijl •!• L ^ 1st  "in  ^1 4-  i i_i -t***«ii\3 jj iiv"i  olIT Go Olr G VGT1  Wfl 1OX1  COTDDOUri 11 *^
^itL *^*'6fit1   Pt'oii^r't'  +-i»   L   i *         *      vsr16cy oi  S3IDPXinif m^d13 wt^r*f^
Cah? SihL. Pateffr.i.J'l. "' !, ?rea^in& the toxic air  pollutants  (TAP's^
     ?ateffori^
       caniJtS?a
                                     air  pollutants   (TAP's)   into
                          or*anico Compounds (VOC's) were  collected
                                 2'4-dinitrophenylhydrazine   (DNPH)
                               compounds (SVbC's) with polyurethane
   . hav
             hP
                                               s.   Future  plans  call
                                           inventory that is currently
                        i-  u~uneS?ne'    dichloromethane.    benzene,
                        ^lchlorodifluoromethane,   4-ethyltoluene  an^
                        to have the highest  ambient concentrations in
                          heric  chemical   fate  of  these  VOC*s
                                8?^ will  focus on this list of VOC's
                                reactions  that  may occur as  well  al
           sub ieo    -e  ,w^h. these   -species.  Due  to  the
        not "he  ii,,rJ   I assiimed that all  the   VOC's  are  primary
                 result  of previous atmospheric reactions.
  n.
coS°^  fullv
   pile aan
         Iff
                           .Envir°™?ntal Management Project  (IEMP)  is an
                       environmental,  study  fe.g   groundwafcer.   radon?
                      i >«?' i determine  routes  of  exposure   to   toxic
               vue  ?L ieyel?   ?fu toxic-  species  in   the various
               s.  The -ir he Potential  human risks from exposure to  these
          r1  identff?^   toxics   monitoring  portion  of   this  effort
           -  The A?r pi??4°n and quantification of toxic species in  the
           .  He          ^t ion Control  Division  APCD  of  I
       f
                       ion Control  Division (APCD) of
                                    this    '
                                 h      r
                                         r  posslble
                                                           Colrad

                                                            impacts
                                was   conducted  during the summer of
                               winter of 1987/88  (November   through

                <3«r"i"-VjttW s^*?sa   ste^inpllng- was  conducted  for  VOC's  using   'Su^rla'
VR  XS1?'*,  and  f^1Sl'erg^ f9,r  SVOC's  with  PUF/XAB-2   r?sin
^ciSeqUen^ °n sili       «ldehydes/ketones   with  cartridges  filled

fcX^»JS.«    L?™-f« -V°fsrl8Bn5°>U-C5oed SfeS-fc
     ^—               !Si and elemental  particulate analysis  (1).

                                   volatile  organic  species   (VOC)
                                 levels  in  the Denver  airshed,   and
                                health effects associated with  these
                          were  routinely   analyzed  are  listed   in
                          lethane,   dichlprodifluoromethane,  benzene,
                              and tetrachloroethene were  among   the
          PreLnt°a

          sPecies
          e,
       oiuene ,di
                           377

-------
highest  concentrations in the Denver airshed.

Experimental Methods

      In   this  project,   sampling  for  VOC's  was   performed using.
polished  stainless  steel  canisters.  The  sampling  units,  provi
EPA-RTP  from the Boise,  Idaho study, also contained sample  ports  *
DNPH   cartridges.   Air is drawn in a heated sample  line by a roughage
From  a manifold,  some of the air is pulled off by  a  metal  bellows
After passing  through  a Tylan mass flow controller,  the pump and a
system,  the air enters the canister. The sample unit  is  also  temp61
controlled to maintain sample consistency.
     During  both  the  summer  and  winter  periods,   24-hour sampl*L
collected,  with the mass  flow  controller  maintaining  the  flow  » iy l>
liters  per  minute.   Initially,  the  canisters  are under approxinia]' , v
inches  of  vacuum.  After sampling, a pressure of   approximately  16  P^n j!
observed.   This  makes  extraction of the samples much easier as no
needed  to  draw a sample of  air  from  the  container  and  there
chance  of  contamination.  When the sample day was  completed,  the
were collected and shipped for analysis.                                -jj

     With  the summer portion, samples were analyzed using  a  GC  wi ^
and  ECD.   Unfortunately  with  this method, benzene and 1,2-dicnlor^  j«f
coelute, as do ra-xylene and p-xylene. The samples  were  not  a?8^^ **?<.
o-xylene   in  the  summer.   The  benzene/1,2-dichloroethane coelutiViuejii'j
problem as benzene was expected to be one of the  major  VOC  const*-/us ,
Thus,   in   the  winter  period,  analysis  was  performed  with  a
eliminate  this problem.  The disadvantage to this  method,  however,
the  detection  limits are higher and some low concentrations w?re,t
missed.    In    these    analyses,    o-xylene    was    determine^
2-chloro-l,3-butadiene  was  eliminated. After the  analyses were Pei
the canisters were cleaned, evacuated and sent back for reuse.

Results

     From  the Denver IEMP,  it was found that  the  concentrations   ,  Or  #
of  the compounds  were  fairly low and often below detection t^ati    ^
detected.  Table 1  includes  a general compilation  of the  concentra,
well  as   how  they relate  to values found in literature.  It shoui.flVe
noted that  the IEMP  concentrations  are  preliminary,   as  they
undergone  final validation.                                           ^ V

     As  a   general  overview  of  the concentrations in Table li   ^erf $
seen that  n-undecane.   dichloromethane,  dichlorodifluoromethane,
tpluene,    4-ethyltoluene,    m-/p-xylene  and  tetrachloroethene
highest  summer/winter  concentrations.


Discussion

     In  looking  at   the  potential chemical reactions that can 0£iJjJj
with these  compounds,  it was  assumed  that  they  are  prii
only.   Also,   due   to  the   complexity of the subject,  only —
air reactions were considered.  The reactions ((2)~(15))  are as
         n-Octane
                Crr i f\tw  ^ f* It  • IT f\
               4t*n  ^"*     J "/<  tljV
                  CgHf£ + 0^ ——^ CjHf.O^

                      a)C,HltO t Ot ~» C4H(HCHO
                      b)CjH140 — •> CiH,^)H
                      cJCiH^O •*- NOi —-»
                  2)c*H«°i * N0  —* c»H
              C,H,7+ NO, ~
                        ! — * C
                              NO — » CgH/40
                              NO, --* CgH^NO,,


                         CtH,f> + Oj. — * HO,. + Cttt,f>
                              + NO — ^CjHuONO                     /BO*'
                                  — * CiH^O + HNO
                              + NO^ — » CjHMONO
                                          +• HONO
                                    378

-------
n-Nonaner n-Decane. n~l)ndecane ---- Reactions are essentially the same as
                                    with n-Octane.

Chloroform
      CttClj + OH --> CC1, + HtO
           CGI, + QL — * CCljOj
           CCljOa. + NO — » CCljO + N0t
           CC130 + Ot — > COCl^ * C102

Dichloromethane. 1 . 1-Dichloroethane. 1.1. 1-Trichloroethane.
      1. 1.2.2-Tetrachloroethane --- Reactions are essentially the same
                                     as with Chloroform.
      Also get: ClCHtCHC10 -f N0t — >COCl,. + CH3NOt after the reaction step
           with Ot with 1, 1-Dichloroethane.

       Tetrachlorid
      CC1,, +OH ~» CCl^OH
           CCl^OH + Ot — » HC1 + ClO^* COClt

      CCl^ + 0 — > CCl^O
           CCUO + OH — > COClt + CHCljO
           CHCltO + NO — » CHCliNOj.

      (very alow)  CC1,, + H — * HC1 + CH,

      CC1,, + hv (X>220nm) — * CC1, > Cl
           CClj + CC1, — > C^Cl,

      CC1,, + hv — > CCl,, + Cl^.
           2CClt — * Cj.01^

Bichlorodlflunrnmethane   Essentially no tropospheric reactions. Those
                               listed below are potential reactions, but
                                proceed  so slowly as to be negligible.

      ClrCFfc -i- 0  — •> COClt +  Ft

      Cl^CF,. + OH — > COClt +  HF  +•  F

-£i£blorof1iinrnmethane  --- Reactions are essentially the same  as
                            with  Dichlorodifluororaethane.
              + OH — * C^HjCHt + HtO                                   (8X)
                   — > C^H^CHjHOH                                     (92*)
           1)C«HUCH,HOH + Ot
                                           .
                             (+0 , +NO) — * CH,COCHO + CHOCH=CHCHO
                     CHOCH=CHCHO (+OH, 0  . NO) —
                    HOH + NO,. — * CC(=0)CHt + HC10
                     with Ethylbenzene
                 OH --» CK,.=CC1 + HZ0
                    — » CH^OHCHCl
                      0t — » CHtOHCOOHCl
           CH^OHCOOHCl + NO ~» CHtOHCOHCl * N0».
                a)CHtOHCOKCl — * HCCO + HC1 * CHO
                b)ChYOHCOHCl + Ofc — * CH»OHCOC1 * HO,
                     CH1OHCOC1 — > HCHO + CO + HC1

                                   379

-------
               CH^CHCl + 0 — » CHtOCHCl
                   CHj.OCHCl — * CHjCOCl


               CHt=CHCl + NOj — * CHj.ONOi.CHCl
                               » NOt + CfyCOCl
                             OtHCl + NO — » CHtONO».COHCl + N0fc
                       CH^ONOtCOHCl — * N0t * HCHO •*• CO + HC1


               CHt=CHCl + 0,. — » CHjOOCHCl
                   CHLOOCHCL — •» HCHO + CO *• HC1


               CHj.=CHCl + 03 — » CHjOOOCHCl
                   CHipOOCHCl — •• HCHO + C0t + HC1                      (24*)
                            — * CH^OO +• CO + HC1                      (76*)


               CHt=CHCl •*• Cl — * {CH,.C1CHC1)*
                          — * (CHj.CHClCl)«
                   IHCHjCHClCl)* — >CHfc=CHCl + Cl
                             ~* CH».CHC1C1
                   2)(CHtClCHCl)* ~»(CHtCHClCl)


          Vinyl idgne chloride.  Trichloroethene.  Tetrachloroethene. g-
               butadiene 	 Reactions are essentially the same ns with
                           Vinyl chloride,

                                                                        f  #
                                                                     Vi  O*  tlr
      From the chemical reactions,  a list of end products for   eac:H   of,.!ut
compounds  can  be developed.  Table 2 is a  small compilation  of |OJD \&$
end   products  that  may  form  from  the   eight   compounds    ol
concentration.                                                            •.$

      Primary  human  health effects and target  organs for the ^^rLjt^.rf'
of highest concentration measured during the Denver IEMP are   P.re?2n  K>'
Table   3.  Of  these  eight  species,  one  (benzene)  is  a  £n;,ro6*'I!ril'
carcinogen and two other species  (dichloromethane  and  tetracluoi   ^p^
are    ppssible   human   carcinogens.   The   remaining  five  sPSS1^ts»   *"
presenting  potential  problems  with  local and  systemic   ef *e?cjty'j.hr
either unclassifiable or had low potential  for   human  carcinogen**- gjj &
general,   the  concentrations   of  all  the  species  are  low  en
chronic effects will probably  be the main area  for concern.

Conclusion                                 /'                            j$
     •"•                                                                a  J*fj|
      From the chemical reactions,  it  can   be  seen  that  there yQG>s$i
potential  reaction  pathways  and many potential end products f°£ pat&$
general,  their reaction  with OH appears to  be the most  important- &i$°
especially  during  the   daytime.   During the night, NO  reactions      *
play  a significant role,  but not with all VOC's.               .  aKed   tfiih
      Concentrations  of   VOC's  measured  in the   Denver    ^^rgr^ai
generally  of  low  levels.  This  will  most likely	1-"i'a r°" —   "
acute  health effects from   any  of  the  species,   and  any      cuua5«
secies   whose  ambient  concentrations  are   below documented JV-j  r OC5
threshold response levels.  What  is  not  clear  is  the
effects   of  the known and  suspected carcinogens in the
of the ongoing goals of  the Denver  IEMP  will   be  to
data  based  and  modeling   based  risk assessments for these
species.

References

1. Air Toxics Monitoring Plan  for Denver IEMP.  US EPA Region VII1'
     Environmental services Div.  March 1987.

2. Graedel,  T.E.. Chemical  Compounds in the Atmosphere. Academic
     New  York. l9'/a, """       		
3. Riggin,  R.M.,  Technical Assistance Document  for
     of  Toxic^Jrganic Compounds in Ambient Air.  US b'PA.

4. Finlayson-Pitts, Barbara J.  and Pitts, James N.,  Jr., AtmogEfe|jgy
     Chemistry:  Fundamentals  and Experimental Techniques,^Tonn
     Sons,  New York. 19tib, luaap.
                                                                . _
5. Weast, Robert C.. ed.. CRC Handbook of Chemistry and PhyjU£3'
     Press,  Boca Raton. 1985:"

6. Chuang,  Shien C. and Bozzelli,  Joseph W. , "Conversion of .

                                        ""Water Vapor>" r""i

                                     380

-------
7.
  ""^D680*!, Ian and Tedler, John.  "Photolysis of Carbon Tetrachloride  in
    tne Presence of Alkanes,"  Int.  J.  of Chem.  Kinet. 14: 1033-45  (1982)

  ~~TrjUigjunental Sciences: Selected Papers,  Systems Applications, San
    *IQ J aelij 1 UUK	nfiMp                	


    Ph«V' Ernesto C., et al.,"t-Dicarbonyl  Yields from the NOx-Air
    p"°>°°xidations of a Series  of Aromatic Hydrocarbons in Air,"
,    ^aviron. Sci. TechnoK 20: 383-387  (1986).
        J» Ernesto C., et al.,  "Yields  of Glyoxal and Methylglyoxal fron
        *Ox-Air Photooxidation  of  Toluene and m- and p-Xylene,  Environ.
t   -^^^Technol^ 18: 981-984 (1984).                            	

    with800* Roger, et al.. "Kinetics of the Reactions of NO  Radicals
    21.  i?r,Series of Aromatic Compounds," Environ.  Sci .Technol.
l2   *A-  1123-1126 (1987).                  	
       •>  Joseph A., et al., "An Outdoor Smog Chamber and Modeling Study
    ZJR 9ljJene--NOx Photooxidation,"  Int.  J.  of Chem. Kinet. 17: 177-

I3.ru        }>
                 S.,  et al.. "Reactions  of Carbon Dioxide with
              —:__Effects of Activated Carbon," Environ. Sci. Technol.
           ' «£ I O
    Stv^*'.et al-> "Photooxidation at  193nra of Cyclooctatetraene and
    4412 n?Qinto Benzene and Acetylene,"  J.  Phvs.  Chem.  89:  4409-
         uy°5).
   I\^t.
    ChiAJ?*  subramonia and Rowland, F.S.,  "Gas-Phase Reactions of Atomic
18 „   Orine with Vinyl Chloride." J. Phvs.  Chem.  89:  3730-3737 (1985).
  *  ^ch  D
    R~1'  -*S'it  et ?!•» Network Design and  Site Exposure Criteria for
         ^g^Nor.criteria Air Pollutants. US  EPA.  H'L'f.
           Marshall,  Handbook of Toxic and  Hazardous Chemicals and
                 2nd ed..  Noves Publications,  fark Hictge.
               iGuide to Chemical Hazards.  US  Dept.  of Health & Human
               ^	-'  n, IJ.C. iy«b,	
                  Values for Chemical Substances  and Physical Agents
                  vironinent with Intended Changes tor l9«3-H4. Amer.
          01  Indust.  Hygenists, Cincinnati.  1983,
                                  TABLE 3


               Health Effects of Significant*  VOC'9 (17}(18){19)
  To1"^
TWA
(ppm)
^_
ne 10Q

oromethane 1000
n«ne 50

10


100

e
100

STEL
(ppm)
— —
500

1250
200

25


150

—
150

Ambient
Cone.
(ppm)
ND-0. 06
ND-0. 69

ND-0. 01
ND-0. 01

ND-0. 08


ND-0. 01

ND-0. 03
ND-0. 04

Carcin.
Effects
Ukn
C

Ukn
C

A


Ukn

Ukn
Ukn

General
Health
Effects
Ukn
Local &
systemic
Minimal
Local fc
systemic
Severe
local t
systemic
Local fc
systemic
Ukn
Local &
systemic
Target
Organs
Ukn
CVS, CNS

	
Liver, eyes,
kidneys, RS
Blood; bone
marrow, CNS

Skin, liver,
kidneys, CNS
Ukn
Skin, liver,
kidneys, CNS
     „        at elevated  levels during the Denver IEMP.


  JVS ^ °8|itr!l[?' un"ur«,  no  information found
                                  381

-------
                              TABLE 1
                     Concentrations of VOC's Analyzed
Summer IEMP
values
Compound (ppb)
n-Octane ND-12
n-Nonane 0.2-14
n-Decane 0.6-165
n-Undecane 0.1-64
Chloroform ND-4
(Trichloromethane)
Dichloromethane 13-685
(Methylene chloride)
1,2-Dichloroethane (2-83)
(Ethylene dichloride)
1 , 1 , 1-Trichloroethane 0.5-40
(Methyl chloroform)
1,1,2,2-Tetrachloroethane ND-2
Carbon tetrachloride 0.1-2
(Tetrachloromethane)
Dichlorodifluoromethane ND-13
(Freon-12)
Trichlorofluoromethane 3-17
(Freon-11)
Vinyl chloride ND-234
(Chloroethylene)
Vinylidene chloride 4-244
(1,1-Dichloroethylene)
Trichloroethene ND-18

Tetrachloroethene 0.3-11
(Perchloroethylene)
2-Chloro-l,3-butadiene ND-15
(Chloroprene)
Benzene (2-83)

Toluene 3-94
(Methyl benzene)
Ethylbenzene 0.6-13

o-Xylene 	
(1,2-Dimethyl benzene)
m-Xylene (2-44)
{ 1,3-Dimethyl benzene)
p-Xylene (2-44)
(1,4-Dimethyl benzene)
Styrene 2-20
(Vinyl benzene)
4-Ethyltoluene 0.5-32
(4-Ethyl-l-methyl benzene)
Chlorobenzene ND-42
Winter IEMP
values
(ppb)
ND
ND
ND
ND-2
ND-3

ND-7

ND

ND-2

ND-1
ND

ND-ll

ND-3

ND

ND

ND

ND-7

	

ND-26

ND-78

ND-3

ND-3

(ND-58)

(ND-58)

ND-24

ND-19

ND
Literature
values (ppb)
(2H3)
0.04-3.4
0.1-9.0
1.0-11.2
0.95-8.8
0.019-5.0
0.004-0.25
0.049-9.4
<0.005
0.045-2.5
<0.005
0.13-2.7
0.03-0.37
0.002-0.077
0.11-2.9
0.0004-0.26
0.38-3.1
0.08-1.0
0.20-1.9
0.05-0.80
<0. 004-0. 22
<0.005


0.005-2.5
0.014-0.35
0.034-7.6
0.010-1.25


0.11-37.0
0.025-57
0.11-65.0
0.005-129
0.05-19.0
0.1-22
<0. 010-9. 7
0.5-33
(0.11-23.0)
1-61
(0.11-23.0)
1-25
1.5-5

0.06-7.4
1.1-13
0.009-2.8
Classification
Alkane
11
il
Alkanic halogen^*1

II
rf

if
it
ti
11 '
it
it
it
M
"
ted
Olefinic halo**1
it
rt
H
M
M
tl
rf
it

Aromatic

11

It

tf

If

If

tl

If
*cd I'0"'1'
Chlorinated
(ND = Not detected or below detection limit)
Compound
n-Undecane
Dichloromethane
Dichlorodifluoromethane
                               TABLE Z
                                               *
                 Potential Fate of Significant  VOC's

                       Potential_End Products_(5X9)(10^.^-

                       Undecanal, undecanone,  undecanoic a  a**
                       Phosgene, hydrochloric  acid,
                       No significant tropospheric  rea
                            potential to  go to phosgene,
                            acid.                 j    *-hon
                       Phosgene, hydrochloric  acid,  carv  ,,
                       Phenol, nitrophenol, resorcinol, »
                       Cresol, nitrocresol, nitrotoluene,
                            benzaldehyde, methyl  gly°"
                       Dimethylbenzaldehyde, glyoxal»
                       Tolualdehyde. dimethyl  phenol, »?
                            dimethylnitrophenol,  glyoxai.
                            glyoxal.

Measured at elevated concentrations during the Denver I£

                              38?
Tetrachloroethene
Benzene
Toluene

4-Ethyltoluene
                                                                  *
                                                                    x

-------
       OF APPLICABLE EMISSION MEASUREMENT
           FOR THE MEASUREMENT OF VOLATILE
        WASTE EMISSIONS
          California
    **S Q —,
         eeult of state and federal regulations, there exists  the  need  to
          docuraent volatile and particulate matter emissions from  both
        ed_ hazardous waste sites and controlled  treatment,  storage,  and
    pe  acilities (TSDFs).  Many measurement technologies have been
& *6 is a* ^este^« and used to assess emissions from these sites.  Litera-
        ai"'-a^l  regarding a variety of volatile  and particulate matter
                   technologies.
    Th'
u  e to i   ^er 3-s an overview of emission measurement technologies  appli-
 °^e  wa 2arc^0us waste site assessment, for both controlled and uncon-
      Ea v,6 s^tes*  Measurement technologies will be presented by  generic
       a   class of measurement technology will be described  in terms of
       ^lication, limitations, and type of data obtained.   Method
            mati°n wil1 allow the reader to select the appropriate
             r°ach for a given application.  Method specific  information
 tn-   metnoc^ description; theory of operation; equipment/supplies
Pte  8ie }ja' ^dvantages/disadvantages; and method references.   Although the
    nte,j ,  "een placed on volatile species emissions, technologies are
          r Particulate matter emission measurement technologies.
                                 383

-------
                 OVERVIEW OF APPLICABLE EMISSION MEASUREMENT
                     TECHNOLOGIES FOR THE MEASUREMENT OF
                     VOLATILE HAZARDOUS WASTE EMISSIONS
         J.A.  Clark
        C.E.  Schmidt
     Radian Corporation
 Sacramento,  California 95825

 Introduction
                                                   T. D'Avanzo
                                        U.S.  Environmental Protection Ag
                                            Boston, Massachusetts 02203
                                                                      nS*
      Many controlled (TSDYs)  and uncontrolled hazardous waste sites c° ..$
 of  "area  sources"  that  have the potential to generate volatile and Par  $
 late  matter air emissions.   The potential health impacts resulting f*°
 emissions from  these area sources can be estimated by determining the
 potential of the site to produce air emissions.   These emission r*
 be  assessed using  a  variety of  technologies, including direct and
 emissions measurement,  ambient  air sampling and  dispersion modeling*
 predictive emissions modeling.   Several  factors  will determine which
 technology or combination of  technologies will be appropriate for
 ating a given site.   These  factors include:  availability of pre-exi0T
 site  information and test data;  the level of effort available to
 these emission  rates;  the level  of accuracy  required for decision'"18
 and the complexity of  the site,  waste material,  and surrounding
                                                                      fl(
     This paper  is  an  overview  of  current  emission measurement/a836  j
technologies  applicable  to  area sources,  including both control!6^
uncontrolled  hazardous waste  sites.   Measurement  technologies will  -h$
presented by  generic class.   Each  class  of measurement/assessment te
gy will be described in  terms of sampling  application,  limitations"
type of data  obtained.   Method-specific  information presented will
the reader to evaluate and  select  the appropriate measurement appr° .
a given application.  Method-specific information will  include:
description;  theory of operation;  equipment/supplies description;
ages/disadvantages; and  method  references.   This  paper  will not
sample collection methods or  the methods for the  analysis  of
     An important distinction to make  in  understanding  the
surement technologies is whether the technology  directly measures    ^
emission rate or whether the technology only measures the reevlt
emission event or gas concentration.   Technologies  that directly
the emission rate typically measure all of  those parameters  that
necessary to calculate an emission rate.  Technologies  that  only
concentration must use an interpretation  (such as dispersion m°
determine the associated emission rate.   Other methods measure one
-------
8' 9  ^faces are the surface  emissions  isolation flux chamber (2. 3, 4, 5,
and the  •     12> 2°' 24' 26>  2?*  28)' head  space samPlers <19-  33. 36),
"lay be  Wlnd tunnel described  by Astle  (1).   Cracks in surface covers also
Waste gfampled ty these technologies.  Vents  at  uncontrolled hazardous
above t^f8 tyPically have minimal or no  flow, and may be sampled by the
(35) ^  ^nnologies, or by head space emissions concentration measurements
ra*es m      measurable gas flow velocities  are  present,  vent emission
      &V be sampled by standard stack sampling  methods (3,  4).
    TK
*s that  pr:Lnc:'-Ple advantage of direct emissions measurement technologies
6B>ission     ^iow the experimenter to calculate,  rather  than model,  and
      n
°Ver man ^^ fr°m measured parameters.  The  experimenter  also has control
^^t jg^ °f these parameters.  The disadvantages are  that  the measurement
1)6 8howtiy affect the emissions event, resulting  in  biased  data,  and it must
S°u*ce.        the rate data collected are representative of the  area

    Th
^e down yo:"atile emission rate for subsurface soils can be measured using
 •' 26»  27     isolation flux chamber (10, 23. 27. 28), soil probes (3.  10.
Sittilar  t '  28^'  and vapor monitoring wells (28).   These technologies are
     (j  b°l     surface emissions isolation flux  chamber technology de-
     stuh",,311'3 °an 8enerate volatile emission rate  data  representative
       S*. 6d"  waste conditions.   When used without "sweep  air," these
    th*»     llelp to measure the soil gas or ambient concentration rather
        emission rate.

    ^-rSSiJEsolation Flux Chamber.    The surface  emissions isolation flux
    of  v , °nf  °^ tne most promising technologies for  the direct  measure-
    cati0 Stile  emissions.   Guidelines have been developed by EPA for
    aPplin  vf  thiS  methodol°8y to land surfaces (20).   The technology is
     e  a ^able to liquid  surfaces (12).   The technology uses a  chamber to
     add H°Wn sur^ace  area  for  emissions measurement.  Clean dry  sweep
     air ^   t? tlle  chamber at a  metered  rate.  Within the  chamber,  the
e*'eep   • m"l-xed with  emitted vapors and gases by the physical design  of
Uer clr inlet  and  an  impeller.   The effect of the sweep air and
 o      cr
te ?etltrati6ates S Eli8ht wind velocity at the emitting surface.  The
atime j!?n °f tne s^naust  gas  is  measured at the chamber outlet by
     is.   struments and/or  is collected as a sample for laboratory
   -«a £    — methods are used  to  determine the number of measurements
  a'*8ed on  aracterize the emissions  from an area source.   These methods
  • Siotl tat     s°urce surface area  and  the variability of the measured
  ieCatl be u  8t rand°mly selected  locations across the site.   The technol-
   ally it,,}6  ^or volatile material,  such as vapors and gases, and is
           ependent of the sampling media used  for sample collection.
   *hft
   .? calcui  lp   advantages of this technology  are that an emission rate
   he _      ed from measurements taken  in the  field without modeling,
            'enter can control the  testing conditions.   The  emissions
             made at the emitting surface,  whether the  surface is a
      --se - * 8n °Pening (crack),  or a  vent.   The technology's principle
   U* *8 not8  t^at em^ss^on rates may be  enhanced or suppressed if the
   'g  uQed t   °Perated according to  the correct  protocol.   The technology
   ptPt°Cess   Satisfy data needs for all  phases  of the site mitigation
    ^ide 
-------
     -                                    t
chamb^^r^^he l^^f ^^^a period of tirae rathe r
and/or gas which builds up m  ^he chamber        F    g ^ ^ operate

the emission flux  for a given  time .   ^  «P   and ^   ic modes.  £  fflf
one of two modes,  herexn ref err ed to as stati       *  emitting Burf»<*
nitrogen.  Emssion           dia  This mode makes

                                      -
                                                                 (
                                       .
                                                            ,
  by the addition of bulk £1« f *= ^ ?he encloaure because  o£ 1

  ^s- ,s.r«^:™ -^« ^ — s thi°ush the so1
  enclosure's bottom edge.                                      clo«u*'


   press the emission rate.



                  Dire. I
    tion flux chamber and can provide

    soils emission rate potential.  These

    sweep air like head space samplers to  ™"™l™ £° &£ the
                                        hnoloies is t
sweep  air like neaa space oom^^ — --        ^^ ^ thflt

tion.  The advantage of using these tec  o og

concentration within the soil pores is higher than


                                386
-------
   Slte*  therefore,  these technologies can provide lower detection limits
ten*1 °tller technologies.   The disadvantage is that a few rather than a
the  S^ntative nuinber °f  subsurface data points are often used to represent
   em^ssions potential of a large subsurface volume of soil.  The soil
    ?pace concentration data can be used as an input to the predictive
        discussed below.

                      Indirect Emission Measurement

    .evetal  indirect emission measurement technologies have been identi-
              concentration profile technology (2, 3, 4. 16, 18), transect
          (2,  3,  4,  16.  18), upwind/downwind technology (16, 18), and mass
   g   (2» 3,  4).  These technologies generally consist of measuring the
   -   ric concentration of the emitted species and then modeling to de-
                    rate.  Many of the models were developed to determine
        concentrations resulting from stack emissions.  For these technol-
          emission source is considered to be a point source.  For area
        Sources,  the source is modeled as a virtual point source.  This
       i} advantage where the overall emission rate for a source, such as a
       is desired.   When the emission rate on a unit area of the source
        fesired»  an additional calculation will be required.  The indirect
            require meteorological monitoring to properly align the
        systems and to reduce the data following sample analysis.

    A H'
 lt ia , 1Sadvantage of indirect emission measurement technologies is that
 C°neen   a  downwind location on the plume centerline.  This technology
   Ptodn  6sted under a variety of waste site conditions and has been shown
       ce valid results.

         "  technology was developed by L.J. Thibodeaux and co-workers at
              of  Arkansas under an EPA contract.  The technology is based
               °f wind velocity, volatile species concentration and
          Profiles in the boundary layer above the waste body.  These
            are used to estimate the vertical flux of the volatile species


  ^o  indicaTPline  e1uiPment consists of a 4-meter mast with  a wind direc-
       at  •  r* wind speed sensors, temperature sensors, and air collection
             •1-08arithmically spaced heights above the area  source; a
                    data collection system; and a thermocouple  for  measur-
                    °r to sample collection, meteorological  conditions
          ^tored to determine compliance with the necessary  meteorological
    4   COJ^T nc® acceptable meteorological conditions are  documented,  the
     
-------
      The advantage of  the technique is in the application in that the
 technology  is  well suited for large area sources and situations with
 laminar  flow over the  area source.   The disadvantages are that the tec
 nology is an indirect  technology requiring modeling to estimate the
 emission rate,  and the technology is labor-intensive and requires a
 amount of field data to be collected to support the model.

                                                                     at
      Transect.    The transect technology measures the concentration
 emitted  species at several locations perpendicular to the plume centete
 This  technology also has been successfully tested at a variety of wae f
 sites.   The transect technology is  an indirect emission measurement *
 proach which has been  used to measure fugitive particulate  and 8afiC
 emissions from area and line  sources (15,  19).  This technique has **
 applied  to landfills,  surface impoundments,  and waste handling opera
 Horizontal and vertical arrays of samplers are used to measure
 tions of  species within the effective cross-section of the  fugitive
 sion  plume.  The volatile species emission rate is then obtained by
 integration of the measured concentrations over the assumed plume a

      The  sampling equipment consists of a central 3.5-meter mast !>*"
 three equally  spaced air sampling probes,  single wind direction, wltfitl>
 speed, and temperature sensors at the top,  and five 1.5-meter masts
 single air sampling probes.   The central mast is aligned with the
 plume centerline.   Two masts  are placed at equal spacings on each
 the central mast and one mast is used to collect air samples at an
 location.  The  spacing of the associated masts is selected  to cover
 expected  horizontal plume cross-section,  as  defined by observation
 profiling with  real-time analyzers.   Prior to sample collection*
 logical parameters  must be monitored to determine if acceptable s
 conditions exist.
                                                                    «ic
     The  transect  is somewhat less  susceptible to changing  meteor    if>
 conditions than  the C-P,  but  does not account for the vertical
of the emitted species due  to  their varying  molecular weights-    iti^
sect, is often the preferred technology  of choice  because  the  re  ^
data can be more useful and the model  can be applied  to a  variety   ^
applications.  The principle disadvantage is that  the technology ^e,
indirect technology, requiring modeling  to estimate the emission
                                                                  &
     Upwind/Downwind .   The upwind/downwind  technology (18)  meaS cjtf>
emitted species at single upwind and downwind locations.   This
has a higher degree of uncertainty than  the  two  previous techno  ^
because of its limited number  of sampling points and  its lack  °
specific model.  It does allow the collection of data at reduce
cost compared to the other technologies.  The upwind/downwind  t
frequently used because it  is  less sophisticated and,  therefore. ^
forward to implement.  The upwind/downwind technology requires  ^
sion modeling to calculate  the emission  rate.  The disadvantage^ ^
technology is that the emission rate data obtained are limits   ^.
ing level and should only be used as a crude estimate of e®i801     .^

     Mass Balance.   A mass balance technology (4)  can be  ufle c
ly determine emission rates.   The mass balance is  a basic  pnn
chemical engineering.  The technology  has been used at disp°SJ^«l*j
where operational data are available,  but its  usefulness f°r
emission estimates at uncontrolled sites was not identified-
technology at an uncontrolled  site, the  concentration of the P
waste lagoon or contained in the landfill waste  would be

                                   388
-------
over yt'.  and  the emission rate would be calculated as the loss of species
""enteri"11116"   The technology,  while useful for an active system with docu-
teaste  and metered inflows and outflows, has disadvantages for uncontrolled
of tirnf^68"   First,  most investigators cannot wait the considerable amount
eec0tl(j  recluired to allow for the measurable losses of material and,
the an',smal1  losses  are difficult to measure because of the precision of
Pte^  a  ytical methods  available.  The complexity of the typical waste
       Qt an  uncontrolled site also could result in a need to collect and
       a large number  of samples, at potentially considerable costs.  Mass
       ls an  approach  that  should be used as a screening technology to
      ^te  emission  rates.

                       Air  Monitoring Technologies
    Air ......
       monitoring technologies  that measure the ambient concentration
           otn area source emissions consist of ambient air monitoring
  ,^e      _combined with air dispersion modeling to determine the area
^asute  lssion rate.   The primary difference between indirect emission
 ich me  t technologies  and  air  monitoring technologies  is the distance at
    neaaBUrements are  made  from  the source.   Indirect  measurements are
        ^1??  f°urce  (usually on  site)  and may enable distinguishing
               source  areas  on  a given site.   Air monitoring is generally
                   distance  downwind  and usually cannot distinguish
               air sources.   Air monitoring typically measures lower
             because  the contaminant  plume  is  subject to  additional air
              e advantage of  air monitoring  technologies is  that often-
    as Ra4..      can be used  to estimate  emissions from area  sources as
           *v other program  needs,  such as  fenceline  ambient air  concen-
(\ e0roloe'a   r Puklic protection monitoring.   The  disadvantage is that
tJl  * air C   conditions, dispersion,  and  air monitoring further downwind
  UC^8 tLC°ncentrations) adversely  affect  the  measurement process,
           reliability of the modeling  estimate of  emission rate.

                    Emissions (Predictive)  Modeling

cai   y °f w S mode^-s have been developed to  predict  emission rates for a
sitg atld 6acKSte B*te types.   These models are almost exclusively theoreti-
            fcodel generally applies only to a  specific  type of  waste
  e  ^B with68 °^ waste sites for which models  are  available  include
  hoation (t Out internal gas generation, landfills  with internal  gas   >
     ^a8oon   -a^^^ co-disposal  sites), aerated  lagoons,  non-aerated  la-
     ant>      with an oil film, open dumps,  waste piles, spills, and land

         pr&j•  #
        ha    lctive  models  require site and waste characterization  data
         "  Q^*^ • ft  -             *
                     from previous investigations.  These data  also  can
        utoed  aculated from information available in the literature,  or
      fsite- Wlt^ 8ome  level of confidence.  The selection of model  input
             Pecific. literature value, or assumed) should be based  on
               °f the decision-making process and the available level  of
           0  Pe^ific data should be used whenever possible to  increase
       ^ted t a?ission  rate estimates.   Two of the most important parame-
           s.   imPlement the predictive models are the waste composition
              n c°efficients for the waste components.
            rate     •
         ic  fat   estimates  based on modeling can be obtained even when no
      Ottl hist0 S  ate availflble,  especially when specific contaminants,
               cal  records,  are the decision-maker's primary concern.

                                 389
-------
With these concepts in mind, it is important that decision-makers to  ^
qualitative understanding of the data used to produce emission es   ^
from predictive modeling rather than simply relying on the emissio
mates as "fact."
     In a recent publication, EPA- s Office  of Air Q^1^^.^
Standards has published a manual on air modeling from TSDFs  &«>•    &
manual recommends various models for TSDF application which  also  c*
applied to uncontrolled area  sources.  Many of  the  air model a  ^
area source  emission  estimates  are listed in Table  i. ^  w^r u^'
tion on application and reference.  It IB important that  the user
  tand  the'advantages  and  disadvantage  of  predictive models  select^
and  the input requirements  of these models.  In general   the a^btair>
predictive modeling is  that an estimate  of  emission rates .canj*  ific
simply by using assumed  data.  The disadvantage is  that  site-spec
detailed information  are  required  to  obtain representative emissi
 data.

 Conclusions
      When selecting an area source emission assessment approach,
 accurate and a precise technology is desirable to allow coupa*^
 Area source emission assessment technology are listed in Table
 with information on application and limitation of these tectmiq

      Based on the information gained, the following statements ^j
 concerning the use of the various sampling technologies. _ When V    ^o
 direct emission measurement technologies, such as the emission  ^^^
 flux chamber, should be used to measure volatile emission rate     ^
 bodies because they offer inherently greater  sensitivity and      ^ t
 ability.  The flux chamber is not well suited to measuring **\. &**
 extremely large waste bodies of varied composition  (large spa *
 ability) or highly agitated liquid surfaces.  The use of heao
 piers should be limited to the collection of  screening-type
                                                                      ^
     For waste bodies where direct emission measurement
appropriate, indirect emission measurement techniques, sucft  ferred-
tration-prof ile technology or the transect technique, are pr  ^^ o
decision to use one or the other technology must, however, oe^      g
type of area source, environmental factors, etc.  When using ^
pling technology, consideration must be given to  the  ability
necessary velocity  (volumetric flow rate) measurement.  The
technology  suffers  from the fact that the emissions  do««fl'   r
small and.  thus, the difference between two large values. J-     ^
emission rate will  typically have a large variability,  vme    g £0t
the mass balance technology should not be used  as the sole
determining an  emission rate from an area source.                  -*.
                                                               • tiS   t"
     Air monitoring technologies can provide area source
However, these  estimates  can be limited  in  accuracy and  pre
meteorological  and  site situational  factors and influences.
"fenceline" air monitoring programs  can  provide useful da
other needs such as community  relations  programs.
                                                                       o .
                                                                        (l
       Predictive emissions models can provide  representativ^
  area source air emissions, especially for TSDF aPPlic®t:L°^er l°
  however, require input obtained from the site source  m
  represent the area source in question.

                                     390
-------
 p  e w°rk  presented  here was  sponsored in part by the U.S. Environmen-
^  r°tection Agency, under contract number 68-02-4392.  It has not been
_    to  Agency  review and,  therefore,  does not necessarily reflect the
   TJ ^° °^^:'-c:i-al  endorsement  should be inferred.  Special thanks to W.
    a^four  and  edited by Susan Penner,  members of the Radian project
                                 391
-------
References

1    Astle, A.D., R.A. Duffee, and A.R. Stankunas.  Estimating _ Vapor
     Odor Emission Rates from Hazardous Waste Sites.  In:  National
     ence on Management of Uncontrolled Hazardous Waste Sites, U.S.
     ronmental Protection Agency, et al.. Washington, D.C.. 1982.  v*
     326-330.
     Balfour. W.D.. R.G. Wetherold. and D.L. Lewis. Evaluation of
     sions from Hazardous Waste Treatment.  Storage, and Disposal *
     ties. EPA 600/2-85/057.  U.S. Environmental  Protection Agency.
     nati. Ohio.  1984.  2 vol.                                            (
                                                                   aCheS
     Balfour. W.D.. C.E. Schmidt,  and  B.M.  Eklund.   Sampling  Appro
     the  Measurement  of Volatile  Compounds  at  Hazardous Waste Sit
     Radian  Corporation. Austin,  Texas, no  date.  29  pp.
                                                                   Ruri1^
     Balfour, W.D.. and C.E.  Schmidt.   Sampling Approaches for Mea^
     Emission Rates from Hazardous Waste  Disposal Facilities.  *»
     Corporation,  Austin.  Texas,  1984.  13  pp.

     Balfour. W.D., B.M. Eklund.  and  S.J. Williamson.   Measurement.0^
     Volatile Organic Emissions from  Subsurface Contaminants.
     Corporation,  Austin.  Texas,  no  date. 20 pp.                     ^
      DeWolf.  G.B..  and R.G.  Wetherold.   Protocols for Calculating.
      sions from Surface Impoundments Using Emission Models:  ie    o2-3
      Note.  Radian Corporation. Austin. Texas. EPA Contract No.    ^
      U.S.  Environmental Protection Agency. Research Triangle ra
      Carolina. 1984.  34 pp.                                         fl
                                                            . ^-« VOC
      DeWolf,  G.B.,  and R.G.  Wetherold.  Protocols fox
      sions from Land Applications Using Emission Mod*
      tion. Austin.  Texas, EPA  Contract No. 68-02-3850, U.S.
      Protection Agency. Research Triangle Park. North Carolina.
 8    Dupont. R.R.  Measurement of Volatile Hazardous
      Journal of  the Air Pollution Control Association.
                                                                  £
 9    Eklund. B.M.. W.D. Balfour. and  C.E. Schmidt.  Measurement^ :
      Volatile Organic Emission Rates.   Environmental  Progress,
      202.  1985.

 10   Eklund. B.  Detection of  Hydrocarbons  in Groundwater by
      Shallow Soil  Gas/Vapor.  Radian  Corporation.  Austin, i

      PP-                                                          , tfP.
                                                              ent
 11   Eklund. B.M., W.D. Balfour.  and  C.E.  Schmidt.   Meas?r^
      Volatile Organic  Compound  Emission Rates with an Emissa-
      Flux Chamber.   For presentation at:  AICHE Summer Nation
      Philadelphia.  Pennsylvania,  1984.  8 pp.

  12  Eklund.  B.M..  M.R.  Kienbusch.  D. Ranum. and T. Harris°g^.f
       of a Sampling Method for Measuring VOC Emissions from
      ments.  Radian Corporation,  Austin, Texas, no date.   Pf    ^p
  13   Farino. W.. P. Spawn, M. Jasinski, and B. Murphy.  R
       AERR Models.  In: Evaluation and Selection of Models
-------
     ir Emissions from Hazardous Waste Treatment, Storage, and Disposal
    Er^-  tieS' Revised Draft Final Report. Contract No. 68-02-3168. U.S.
     nvxronmental Agency, Office of Solid Waste, Land Disposal Branch.
    1983.  pp.  5-1 - 5-13.
          .  W.J.,  M.S.  Yang,  J. Letey. W.F; Spencer, and M.H. Roulier.
    in s .   posal  of Hexachlorobenzene Wastes: Controlling Vapor Movement
    Po   £     In:  Land  Bisposal of Hazardous Wastes. Proceedings of the
    ARG   Annual  Research Symposium,  U.S.  Environmental Protection
     8 ncy.  Municipal Environmental Research Laboratory. San Antonio.
     exas. March 6,  7,  and 8.  1978. pp.  182-190.

    parm
    Hexa^i  M*J"  M'S'  Yan8'  J- Letey, and  W.F.  Spencer.  Land Disposal of
    EPA6on/°r°benZene Wastes:   Controlling  Vapor  Movement in Soil.
    Rese      ~119'  U'S'  Environmental Protection Agency, Office of
        arch and Development,  Municipal  Environmental Research Laboratory.
     •^cirmati.  Ohio. 1980.   6Q nn.
                    >      t
16
    Sour8'  S'T* MeasurinS  Rates  of Volatile  Emissions  from Non-Point
    Meet^e  Hazard°us Waste Facilities.   Presented at  the 75th Annual
    ana  ^fo^ the Mr Pollution Control Association,  New Orleans,  Louisi-
      •  19
               22 pp.
      no  em   „
    from L       Estimating and Field-Validating  Hazardous Air Emissions
    Cotifft   d DisP°sal Facilities.  In: Third  Pacific  Chemical Engineering
       erence, Seoul. Korea. 1983. pp. 338-343.


    Ptofr?' S'T"  Model Prediction of Volatile Emissions. Environmental
      gtess 4(2): 141-144. 1985.

     api •!
         8'  E*J'» A-J- Kurtz, and M. Rahimi.  VOC Sampling for Emission
        ntH          Snd Ambient Air Quality on an Inactive Landfill.
        I9g  Soil-Gas Measurement for Detection of Groundwater
             °n by Volatile Organic Compounds.  Environmental Science
             ogy,  2 1(10): 1022-1024. 1987.
           D
            '*  Snd  £>"J>  Leinonen'   Rate of Evaporation of Low-Solubility
             8 from Water Bodies  to Atmosphere.   Environmental Science
             ogy.  9(i3).  1178-1180.  1975.
       a

   S°^ V°iatii-and.A*T*K*  Yeun*   Mass  Transfer Coefficient Correlations
,    CieQce  o ^lzation  of Organic  Solutes  from Water.   Environmental
?t          and Technology,  17 (4) :211-217.  1983.

      ian c
3.  CSl Memorl^ration>  3M Foam Evaluation for Vapor  Mitigation:   Techni-
i          «ndum.  Sacramento, California.  1986.   95 pp.
    ^ian
                       Survey ^d Assessment  of Air  Emission Modeling
                 Landfills. Draft Final Report.  Research Triangle  Park.
                 1988.  115 pp.

                                393
-------
26   Schmidt, C.E., and J.K. Meyer-Schmidt. Assessment, Monitoring, an ^
     Modeling From a Superfund  Site Remedial Action.   Presented at the
     Pollution Control Association Annual  Conference,  San  Francisco,
     California,  1985.  20 pp.

27   Schmidt, C.E., and W.D. Balfour.  Direct Gas Emission Measurement
     Techniques and the Utilization of Emissions Data  from Hazardous
     Sites.  Reprinted from:  National  Conference  on Environmental
     neering Proceedings, Environmental Engineering Division,  ASCE,
     8 pp.

28   Schmidt, C.E., R. Vandervort, and  W.D. Balfour.  Technical Appr°
     and Sampling Techniques Used to Detect and Map  Subsurface  .
     Contamination.  For presentation  at  the 79th Annual Meeting, **  ^,
     Pollution Control Association, Minneapolis, Minnesota,  1986.

29   Shen, T.T.  Estimating Hazardous  Air Emissions  from Disposal Si
     Pollution Engineering, 13(8):31-34,  1981.

30   Shen, T.T., and G.H.  Sewell.  Air Pollution Problems  of
     Hazardous Waste Sites. Civil Engineering for Practicing and
     Engineers, 3(3):241-252, 1984.

31   Shen, T.T.  Air Pollution Assessment of Toxic Emissions from
     Waste Lagoons and Landfills.  Environment International,
     1985.

32   Shen, T.T.  Air Quality Assessment for Land Disposal  of
     Wastes.  Environmental Management, 6(4):297-305,  1982.

33.  South Coast Air Quality Management District.  Landfill  Gas
     Report of the Task Force, El Monte,  CA, 1982.
                                                                 toeJlt'ft
34   U.S. Environmental Protection Agency.  Hazardous Waste,  Tre    Wf
     Storage and Disposal  Facilities (TSDF) — Air Emission  Mode  '  ^pfr
     Report, Office of Air Quality Planning and Standards, Resea
     gle Park, North Carolina, 1987.   pp. 6-1 to 6-16.                .
                                                                  ati"  rf
35   Wetherold, R.G., and D.A. Dubose.  A Review of  Selected The^  ad
     Models for Estimating and Describing Atmospheric
     Disposal Operations.  EPA Contract 68-03-3038,  U.S.
     Protection Agency, Office of Research and Development,  J-110""^
     Environmental Research Laboratory, Cincinnati,  Ohio,  1982'

36   Wood, J.A., and M.L.  Porter.  Hazardous Pollutants in clafl*  C
     fills. South Coast Air Quality Management District, El  M°°
     nia, 1986.
                                   394
-------
         Predictive Models for Estimating Emissions From Area Sources
sheB
      ux a
^Oftav. •
            lent
         Landfill
         "Add On"
                           Application
                          Comment References
                        Covered Landfill
                        Covered Landfill
                        Covered Landfill
Covered Landfill
Covered Landfill
Covered Landfill
with Co-Disposal
                        Covered Landfill
                        with Co-Disposal
                         and Barometric
                             Pumping

                        Covered Landfill
                        with Co-Disposal
                          Open Landfill
                          Open Landfill
                         Open Landfill,
                             Spills

                         Open Landfills.
                           Waste Pile,
                         Land Treatment
Based on Pick's First Law
of steady state diffusion
(14. 15. 25)

Modification of Farmer Model
to include weight fraction
of constituent of concern.
Assume Raoult's Law. (13.
25. 29. 30. 31)

Two-resistance Theory Model
(7. 25. 35)

Modern version of Farmer
Model accounting for high
volatile concentrations.
(25)

Based on Farmer Model.
Accounts for Barometric
Pumping and Decline in
Emission Rate with Time
(34)

Accounts for bio-gas
production and flow
through soil (13. 25)

Accounts for bio-gas pro-
duction, flow through
soil and effect of baro-
metric pumping (13. 25)

Accounts for bio-gas
production and flow
through soil (25)

Based on Fick's Second Law
of unsteady state diffusion

Modification of Arnold
Model,  includes wind speed
variable (25. 29. 30, 31)

Modification of Shen Model
using ideal gas law (25. 34)

Applicable to waste mixed
with soil (34)
                                  395
-------
                           Table  I.   (Continued)
 Predictive Model
Thibode aux-Hwang
AP-42 Fugitive Dust
MacKay and Leinonen
Thibodeaux, Parker,
and Heck
Smith, Bomberger,
and Haynes

Shen
RTI
    Application
    Comment References
  Land Treatment
Covered Landfills,
  Open Landfills
Non-Aerated Surface
      Impound
 Aerated and Non-
  Aerated Surface
      Impound

    Non-Aerated
  Surface Impound

    Non-Aerated
  Surface Impound

 Aerated and Non-
  Aerated Surface
     Impounds
Applicable to waste
with soil (35)

Empirical equations »
on silt content, vehic
travel, etc.  (AP-42)
Non-steady state mo
based on two-film r
tance theory (6, 22
35)

Based on two-fil°
tance theory (2. '»
                    35)
For highly volatile
species only (35)

Empirical equation   .j
used for screening
Based on two-res
theory, can accoa°^
biodegradation
  This table is not inclusive of all available predictive
                                   396
-------
      Table  II.   Application of  Emission Rate Measurement  Sampling
                 Technologies to Area Emission Sources
Surfpi
»C£  Is°l"iona
 Active  landfills
                      Inactive controlled
                      and uncontrolled
                      inactive landfills
                                                    Limitations/Comments
                                     Small cells  with uni-
                                     form waste composition

                                     Can be used  on surface
                                     and for vents  at land-
                                     fills
         Sampl
,ers
                      Surface impoundments
                      and lagoons

                      Landtreatment
Applications similar
to the surface
isolation flux chamber
                      Inactive  controlled and
                      uncontrolled  landfills
                      Surface impoundments
                     Waste piles
                     Inactive controlled and
                     uncontrolled landfills
                     Subsurface contamination
                     Surface impoundments,
                     landtreatment
 Little  or no surface
 agitation

 Generally uniform waste
 composition, applicable
 to large cells

 Typical use is for
 concentration measure-
 ments, data used for
 relative comparison
 purposes

 Used to estimate emis-
 sions under simulated
 wind flow

 Can be difficult to
 perform sampling

 Estimates  of particu-
 late matter emissions

 Used  to  measure  soil
 gas  concentration or
 emission rates subsur-
 face

 Typically used to iden-
 tify  and map  subsurface
 contaminants via  soil
 gas concentration, can
 be used to estimate
 emissions from dis-
 turbed waste condi-
 tions.

Meteorological condi-
tions must meet crite-
ria; not suited for
small impoundments or
landtreatment plots
                                397
-------
   Area  Source
Emission Assessment
                            Table  II.  (Continued)
   Area  Source
Transect   ,
         a D
Technique '
Upwind/Downwind
Technique  '
Mass Balance
Technique
Air Monitoring/
Modeling
Technologies
            a.b
Predictive Modeling
Active landfills,
surface  impoundments,
landtreatment,
drum storage
All TSDF facilities and
uncontrolled waste sites
Most TSDF facilities
                        Uncontrolled  landfills
                        and lagoons
TSDF or hazardous
waste site fenceline
Most TSDFs,
uncontrolled land-
fills and lagoons
  Limitations/CSSSS
Meteorological c°D.(g'
tions must meet
ria; quires mJ"iBiaA
interferences
other upwind

Emission estimate_,
ited technique
ly used as £
technique in tn^
opment of a
more accurately
sent emissions

Must identify fl
capable of
all streams
                            Must space
                            ments  over
                            use detection
                            niques  capable
                            detecting
                            tive changes
                                                                    d
                                                                   lt
tions. terrain.   e
upwind inter^  tjr
will affect ut
analytical se
tivity i
limiting
Models
specific
be represen
o
,  Volatile Species Emissions
  Particulate Matter Emissions
TSDFs - Treatment, storage, and disposal  facilities
                                   398
-------
       AIR   MONITORING
      AT  WASTE   SITES

        BARBOZA,  P.K.
          Jersey  07653
                 THOMAS T. SHEN Ph.D
                 Division of Environmental Sciences
                 Columbia University
                 80 Haven Avenue
                 New York, New York 10032
to
      •
             is  directed to  those who  are  involved with using
                and who are  not well versed  in the  designing
                   air quality monitoring for toxic volatile
         unde  (VOCs) at waste sites.  Due  to the complexities
             sampling  and monitoring  programs often  can  be
             consuming, and may yield less  than satisfactory
       this  paper is on the  value  of  preliminary
        monitoring  program  and  understanding  the
     required to develop  a  successful  program.   A
a  approach is presented in which the objectives  of
A e set by answering  basic questions what,  why, and
  project-specific monitoring plan can be  developed
    by asking when,   where,  how,  who and how much.
    udee  a  case  study  of  air  quality  sampling
    a waste disposal site.
                            399
-------
 INTRODUCTION

      Hazardous  waste  sites pose a challenge   for  ambient all
 monitoring.   The  types  of air emissions  present  may   be  quit!
 diverse  and commonly consist of mixtures  of various  individual
 compounds.   Sampling  can  be  further complicated  due  to an;
 treatment  or  degradation of compounds over time.   Mixtures  o:
 compounds may vary from day to day and ambient  concentrations oi
 compounds  may vary over shorter time periods due  to changes  ii
 site activities, emissions, and environmental conditions.

      Air  sampling  is used to identify the actual  air  quality
 conditions   at a site during a given time  period.   The  need for
 sampling   to  determine potential risk,   via the   air   pathway
 posed by a  site can be triggered by modeling studies. Monitorinj
 results  are  used  to verify air toxic  levels  predicted  wit:
 conservative  emission  (vaporization,   etc.)  and   dispereio:
 modeling or  expected based on  other available  information  abou'
 a  site.

      Initial   program   planning  is  important   so  that   tht
 limitations  of the program are identified  and the  usefulness  o:
 informativeness  of the results can be clearly understood,  Thil,
 paper  discusses regulatory requirements for air monitoring an:
 highlights   the  importance of adequate planning   of  monitorinj
 programs for  volatile organic compounds  (VOCs) .    Emphasis  is
 given  to a decision process of program planning  as   shown  i:
 Figure   1.   A case study of air monitoring planning  is  presents:
 to  illustrate major decision elements of the decision process.

 REGULATORY REQUIREMENTS

      Air pollution concerns at waste sites have received minim
 attention under RCRA  Subtitle C regulations.   No control has ye:
 been  established   for problems associated with  VOC    emission;
 from  hazardous waste sites.   Nevertheless methods of   measuring
 toxic   volatile organic compounds  have been and continue  to b
 developed by  the EPA as  well as  many  state,   industrial,   an:
 academic research  organizations (1,2,3).

      Many states  have adopted air toxics  regulations that  dea!
 with  emissions and acceptable ambient concentration  levels o;
 various    VOCs.    Most  regulations  do  not  include  specif!
 monitoring   requirements  and  when  required is usually   specific:
 on  a case by case basis.   There  is  a  need and a demand for  all
 monitoring   of  VOCs to identify toxic  levels of contaminants a'
 sites   for   health and safety  concerns  and  to  assess off-sit!
 impacts  to protect the public  from VOC emissions.

PROGRAM  PLANNING

     A    decision   approach  for   planning  VOC  monitoring   if
 illustrated   as  a simplified  decision tree in   Figure   1.  Thif
approach  has been adapted  from the  methodology  for   designiiij
and  conducting air monitoring  programs  for toxics  at  hazardoui
waste   facilities   (4).  The  approach  can  be  divided into  thre-
                              400
-------
""Mo*.
    Cc"nponent parte ;

     ~ Objectives
       Planning
       Implementation
                Each of these components is important  to  the
         the program.  However, probably the most important to
    u ,   because it is moat commonly overlooked is identifying
,°oapi? f;rstanding   the  program  objectives.   This  can   be
i* cond    •  by answerinf basic questions.  The reason (WHY  ?)
}6lt*ClttantUOting  tne  monitoring program is  one  of   the  most
Vn Ue^al]  questions to be answered beforehand. The eventual goal
ti e in |K to  determine the existence,  magnitude, and extent of
      the air.   The trend in monitoring requirements is toward
   tovecion of the Public health andwelfare,  from  ambient
           (WHAT ?) in a manner similar to that  traditionally
          witft criteria  pollutants.  However, the measurement
iff of »<*8  *° be mucn raore complex,  because of  the  large
Of  the f ifferent VOCs, the lack of compound specific sensors,
   "^fty   v1' tnat many contaminant situations involve  mixtures
    (HHrm  mical compounds.  Data requirements of various  end
          ?).may dif^r widely.   Regulatory  agencies,  health
       '   8cientists,   and engineers concerned with  design of
          alternatives,   have  different information needs and
   J^NNU
               Once   "the   program   objectives   have   been
                w©  can proceed to the second essential part of
               Planning  or  design  of  the  program  can  be
     1.  TM   by answerinS more specific questions as shown  in
      -that    part of the Procees focuses on the details of the
           all°w the objectives  to be achieved.  Questions to
       (     include  WHERE ? and WHEN ? to sample,  the need for
       x  U"ATION   ?),   and the expected  concentration  ranges
          t^newers "to  these questions help to determine HOW  ?
          t"e  samples  and what types of equipment are  needed .
    *  aiv+P"1'ins   methods may be required  when  sampling  for
       tii Urea or compounds with different properties  (i.e.,
         8ea^'  Becoming   more  available in this  regard  are
   v    aft0    ^   "tubes.   Other  important  concerns  include;
   itio«l  ^8"Jance/quality   control  (QA/QC),   sampling   and
    8>  anSethodologies,   minimum  detection limits,   number of
       na cost.

  '?? imp]   ATION- Additional questions must be considered for
       anal  ntatlon-   WHO ? wil1 Perform the sampling and WHO
            6  tlle  aaraPles.   Laboratory  selection  is  not  a

        ia     "umber of qualified  laboratories to perform air
      6. UQui8lna ^ and turn  around times  tend to be  less  than
      ?) is     H ? wil1 the  Pro«rani  cost (What is one willing
      **» imr,? Con8ideration which can carry significant weight
      iveiy  enentation.    Cost  benefits can usually,   only be
       6 oLWSaeured by the number  of samples and the  level of
            feels comfortable with.

             c°8^e can vary  significantly,   and can range  up
              red dollars  per sample not including costs  for
              design, sample collection,  and presentation and
            °f results.  Due to the  expense,   judgements  are
                             401
-------
often  based   on  the  results of  only a few  samples.  For  thit
reason,  use   of   statistical  analysis  and  identification ol
confidence  intervals  is  rare.

     As  the  decision process continues the more  subtle  issue?
are  addressed,   thus defining  the scope and  content  of  tht
monitoring  program,   as   well  as  identifying  some  of   th*
limitations   of   the  designed program.   A  monitoring  plan  for
review  by various  interested parties should describe the  scope
and  content  of the program.   This process can serve to minimize
misunderstandings  later.   In order to increase the chances of a
successful program,   a consistent approach must be utilised  for
designing  . and   conducting  ambient  monitoring  programs   for
evaluating ambient  levels  of VOCs  at hazardous waste sites.

CASE STUDY
                  ^
     A  24  acre   site  used from the 1930s  to  the  1970s  for
landfilling   industrial  process   wastes  (i.e.,   sodium  salts,
barium,  calcium,  and  sodium chlorides,  contaminated discarded
cell  rubble,  a  variety   of  chlorocarbons 'and  organics  and
inorganic  wastes)  was  closed in 1977  and covered with  a  clay
cap. A ground  water protection program  was implemented to remove
and   treat   contaminated   groundwater,    and  reduce   off-site
transport  of  contaminants  via   groundwater.   As  part  of an
endangerment   study,   a  review (including modeling) of  the  air
pathway   raised  some  concern  regarding  the  potential   for
volatization   and  subsequent off-site transport  of  airborne
contaminants.  The  conservative  assessment  indicated  modeling
projections   of  off-site   airborne hazards  with  concentratior.
estimates  below the  respective TLV' s but at or near  respective
acceptable ambient guideline levels,   for some of the  compounds
of concern. Due to the uncertainty  associated with the modeling
process   more  valid risk  estimates   for  potential  airborne
exposures  were developed  with specific monitoring  of  airborne
contaminants   was conducted.   A plan was prepared describing the
program.

WHAT  ?  The  results  of  a  relative ranking of potential air  ha-
zards indicated that  the highest  potential airborne hazards were
associated with hexachlorobutadiene,  vinyl chloride, 1,1 dichlo-
roethylene,   carbon   tetrachloride,   1,1,2,2  tetrachloroethane,
chloroform, tetrachloroethylene,  and trichloroethylene.

WHY ?  The objective  of  the sampling program was  to identify the
potential  for air contaminants to originate from  the  site.  «
major  concern  was   with   regard to  interferences  from  other
sources in the highly industrialized area.

WHOM ?  The results were to be used by  the client and regulatory
agency  to verify the potential airborne risk estimated  in  the
endangerment asssessment study, by actual field measurement.

WHERE  ?  Five  sampling   locations were  established  including
concurrent   upwind   and   downwind sampling  in  an  attempt to
isolate  and   identify  emissions from  the site  and  any  other
potential off-site upwind  sources.   Typically,  sampling would h
performed  at  off-site  receptors.   However,   in this  case nc
receptors were located nearby,  and because of the expected  lov
                               402
-------
                                 and naar  th«   downwind
^^r** sufficient             historic  (DURATION?)  data was









1 (.  J rt .*«_

                        S£ *he+falVaropling event  are   shown
                        the most part below detectable limits
                         site sources of some contaminants  of
                        e results presented for vinyl chloride
                         conr-i vm^H V,™  riwo   j.. .          . -^^^
     was
                 '        esus
                                 by  OMB-  due  *
       d
       CONCLUSIONS

                                                  of  VOCS  to
                                                 expensive.
             ,  planning and implementation
ig«in
                        °f   Methods   ^r  Determination  of
                  P n!  ln.Ambient   Air.   EPA-600/4-84-041,
                  Protection   Agency, RTF, N.C. .April  1984!

                    Assi8tarjce Documents  for   Sampling and

                                                      "
                                 ^ew Jersey Department  of
           r»,  'Jection"  Hazardous Waste Programs,  Bureau
           tai  Measurement and Quality Assurance, Trenton,

           1988'          403
-------
4.  Asoian.M.J.,  M.J.Barbosa,   and L.M.  Militana. A
    Methodology   for  Designing  and  Conducting  Ambient
    Monitoring at Hazardous Waste Facilities.  APCA, 79th
    Meeting, Minneapolis, Minn, June 1986.
Table 1 MAXIMUM OBSERVED AIR CONCENTRATIONS AND TLV AND
Compound
  Maximum Concentration    TLV
Upwind  On-site Downwind (3-hour)
(ug/m3) (ug/rti3) (ug/m3)   (ug/m3)

Carbon Tetrachloride
Chloroform
1 , 1-Dichloroethylene
Hexachlorobutadiene
1 1 2 2Tet rach loroethane
Tetrachloroethylene
Trichloroethylene
Vinyl Chloride
NB<6,
>404.
65,
ND<45
ND<6.
ND<6.
116.
ND<1
3
Ib
1

8
8
4

ND<
ND<
ND<
ND<
ND<
6
6
6

.
t
7
7
7
44
6
.
ND<6.

ND<
6
.
7
7
7
.93
ND<6
8
ND<6
,9
.8
.9
30,
50,
20,
ND<43
ND<6
ND<6
ND<6
ND<.
.9
,9
,9
83
7,
335,
270,
10,
000
000
000
240
000
000
000
000
                                                          'A
                                                            tf
                                                          n'S
NOTES:(a) ACGIH Threshold Limit Values (TLV),New York AcceP
          Ambient Levels (AAL)
      (b) Compound saturated and not confirmed
          ND = Non-detectable (less than detection
          NA = Not analysed for
                                                           ,t*p
                              404
-------
                            Objectives
O
CJ1
Pbter
VOC x 	 "
I \ Mon-oota
WHAT I SEM-TOC
1 ws/wo
1 METEOR /Temp
1 OLOGY/'
VSabarty
\HumidhY
Ltg*l Action
COMMJANCEy
\Soure*
WHY HEALTH AND SAFETY
Some*
INVEST! -/Scrwnina
QATION/''
VRi*
\Trropoft
REGULATORY
AGENCIES
WHOM HEALTH OFFICERS
SCIENTISTS A
ENGINEERS

/ /
' V -X
WHERE
O
WHEN
V 	 .) DURATION
O '
x 	 x LEVELS
c_^
INSTAN
Q 	 J HOW

UPWMD
: Source

VAmWent
OOWHWWO
SHORT Hour
TERM / Day
\ Week
LONG Smenl
TERM y''
NvAnnual
Re^Time
Hiftortc
Accurtcy

Specif icitv
Survey Metm
TANEOUS / 	
\Tube»

TINUAL /Tuhet/B^.
\G-C.
INTER - Tubn
GRATED /BM>
\Medii
	 Implementation
i r 	 	 ,
OFWd
WHO Vlb (^
^. — _^ »» r\\j 	
C j F^d
CONIRACI/ X" X
O'V* ( )
—
OHOW MUCH COST/ Hl* s«mpi»#
BENEFIT / , 	 \
	 	 C r i
X'-'**S«mrte#V^ }
O
o
o
o
o
                                 Figure 1    AIR MONITORING PROGRAM DECISION TREE


                                            FOR VOCs AT WASTE SITES
-------
PRELIMINARY EVALUATION OF TEST METHODS FOR VOLATILE
ORGANICS IN HAZARDOUS WASTE
Sam B. Balik, A. R. Gholson, and R. K, M. Jayanty
Research Triangle Institute
Research Triangle Park, NC  27709

G. D. McAlister and R. T. Harrison
U.S. Environmental Protection Agency
Research Triangle Park, NC  27709


     Four analytical methods were evaluated for determining the vo *
emission potential of hazardous waste.  The methods tested were
headspace analysis, elevated temperature purge and trap, steam
distillation, and gravimetric purge and trap.

                                                               x
     Each method was evaluated by preparing and testing synthetic
that approximated the chemical and physical properties of six dif'e
waste categories.
     In the evaluation it was found that steam distillation and
gravimetric purge and trap had the highest removal and recovery
efficiencies.  These two candidate test methods also had the best
precision, calculated by triplicate analysis of each synthetic was  ft
equilibrium headspace and the elevated temperature purge and trap *
found to be the easiest and least expensive to perform.

                                                               i afl tl>e
     The results of the method evaluations are discussed as well
advantages and disadvantages of each method.
                                   406
-------
   legislation has been passed to protect public health from the emission
  J? fttlle organic compounds from hazardous waste treatment, storage,  and
       facilities.  As an Important part of any regulatory development, a
       test method must be found.

   The purposes of the test method include assessing both the air
       potential and the air emission rate of volatile organics  in
of v j°Us waste.  Other potential purposes include determining the quantity
  ce       or&anlcs that can be recovered by a commercial  treatment
   8s and evaluating the performance of treatment systems.  These
   8es are directly related to ways the test method can be used  to
 cl]f    a*r standards at hazardous waste treatment storage and disposal

       two most important features of a suitable test method to be used
          implementation are  (1) the ability to remove and recover
     e organics from a variety of waste forms and  (2) the precision of
           Tne test "»ethod should also be cost effective and simple to
   Th
       *est methods currently under evaluation are equilibrium headspace
   gj,  '  elevated-temperature purge and trap  (ETP&T), steam distillation,
    avimetric purge and trap (GP&T).  Evaluation of these candidate test
   u      being performed under controlled conditions by testing six
   a^p waste types that contain known amounts of nine organic compounds.
            waste forms are dilute aqueous, two-phase aqueous, organic,
 " leth    ~3' aqueous sludge, and dry solid.  The nine organic compounds
^Ueh»    ne chloride, 2-butanone, 1-butanol, isooctane, pyridine,
       chlorobenzene, phenol, and naphthalene.  These compounds represent
           of solubilities and volatilities in the various waste
         The compounds were not chosen to represent any particular
      a$ waste type.

           Methods

is*  f$r° 8tock solutions were prepared with the nine organic compounds.
    ln.  Wa8 a concentrated neat solution that contained 4 g of each
   *Qth   "ltf seconc' was a 1* solution of the nine compounds in methanol.
   n_  anol was used as a carrier to prepare the dilute aqueous waste
         the low water solubility of isooctane and naphthalene.  The
        waste forms were prepared by spiking the following matrices with
   -«a  8olution:  organic,  light mineral oil; solid, fuller's earth;
 "ftfaj 8ll*dge, fuller's earth and deionlzed water; organic sludge, light
      °*1 and fuller's earth.  The mixed-phase aqueous waste contained
    v excess of the nine compounds that their solubilities in water were
    »dfX°eededl   Relative concentrations of the compounds in the six
      trices are shown in Table I.
   B6.....
           automated systems are commercially available for the
          headapace analysis and ETP&T, and because these two techniques
          standard methods,  the specific details of their operation were
        However, the steam distillation and gravimetric purge and trap
         are 2 more obscure  and therefore are covered in greater detail.
                                 407
-------
     The headspace analysis was conducted by sampling and analyzing
vapor space above 10.0 g of each synthetic waste contained in a
125-mL vial.  Headspace samples were collected at 25, 50, 75, and 90
     In the ETP&T, nitrogen was purged through the synthetic waste
that were suspended in water to strip out the volatile organics.  The ^
volatiles were removed from the gas stream with a charcoal trap, «hic ^
subsequently desorbed with a 10% acetone and carbon disulfide mixture
analyzed.  The samples were purged sequentially at 25 and 90 °C.  The
charcoal trap was changed when the temperature was elevated.
                                                                      tl"'
     The synthetic waste mixtures were poured into a 3-necked distil  ed
flask for the steam distillation test.  Additionally, 125 mL of dei<>»
water was added to the organic, organic sludge, and solid waste tyPe  ,
because they had no native aqueous phase.  This mixture was stirre d _agjt
made basic with ION NaOH until a pH of 12 to 13 was achieved.  The I ^ jd
was sealed and a heating mantle was used to bring the temperature UP ^
°C.  The vapors were condensed in a two-stage condenser system.   ^iigt**
stage was an ordinary ice-water-cooled Leibig condenser.  The disti  ^t
were collected in an ice-water-cooled receiving flask.  Those vapors ^
did not condense went to the second-stage condenser, which was a dry
acetone cold finger.  Vapors that did not condense at this stage we   ^
collected in a Tedlar® bag.  Sample distillate fractions were col!^led
after 5, 10, and 20* of the original volume of the waste had distil   ^
over.  The distillation flask was allowed to cool after collection    ^F
20% bol lover fraction and the remaining waste mixture was made aci <*  . tl>
50* sulfuric acid to a pH between 2 and 3.  The flask was resealed    ^
distillation was allowed to proceed until an additional 20* of the ^
volume of the waste was collected, for a total of 40% boilover.    ^$
contents of the apparatus were purged with 10 L of nitrogen to re»o ^
lingering organic vapors.  Finally, the condensers were rinsed " *
organic solvent to remove any components that might remain on the
surfaces.
                                                                  ded1"
     In the GP&T, 10.0 g of each synthetic waste sample was suspe"   ed •
100 mL of dioctyl phthalate (OOP).  The volatile organics were strg gtr«{||
from this mixture with nitrogen at a flowrate of 15 L/min.  The
along with the volatile organics, passed through the coalescing
filter contained a porous glassfrit that acted as an aerosol tr
volatiles were removed from the gas stream in  the primary and s*CfCO&
adsorption tubes, which contained  4.0 and 2.0 g of activated cha£aj
respectively.  The trapped organics were extracted  from the cnarcto c
25* solution of acetone in carbon disulfide.  The final step was ^
and analyze a sample of the OOP after the purge to  ensure all nltl
compounds were completely purged from it.
     The percent recovery  of  the  nine  compounds  in all  sample ^^jtf1
generated in this study were  analyzed  by  gas  chromatography (GC'  $ I"
ionization detection  (FID).   Nine component  standards were Prepaart)OP d
deionized water with  a methanol carrier,  carbon  tetrachloride,  c  epar
disulfide/acetone solutions,  and  OOP,   A  gas  standard in N2 was P^j
with the seven liquid phase compounds  in  a passivated stainless  ^A
canister.  Because phenol  and naphthalene are solids at room *    $e
gas standards were not prepared with them.   Control samples were
from activated charcoal to determine extraction  efficiencies fo
trap methods.  Multipoint  calibrations with  the  nine component
were performed to ensure linearity in  the range  of the  collected
                                    408
-------
*6fe   8>  Average  response factors in units of nanograms per area count
eac-  Calculated  for  each  component in every standard.  Percent recovery of
the  °onpound was  calculated by multiplying the sample area response and
c e> aQueous  response  factors were used to calculate the respective
   ent recoveries.

"e8uit»
    ts and Discussion

                 Percent Recovery of the Test Methods

       waste forms were  analyzed in triplicate by the four candidate test
      and an average percent  recovery was calculated for each component.
   Vtt^°Veries of  the nine compounds were averaged to get the total percent
       from the  six waste  forms.   These averaged recoveries are shown in
        tnat the recovery  of  the  nine  compounds depends on both the test
            and tne waste  f°rm  tested.   For example,  the steam
   aixatlon usually has the highest  recovery of the nine compounds from
      Waste forms, especially in  the aqueous-phase wastes.  The lower
       fr°™ t'le orBanic wastes  may  be  due to the compounds'  high
           into the light  mineral oil.   Once in mineral oil, the
        may not be effectively  stripped from it.
           evidence of  this  partitioning into the mineral oil was
   c0fc     the equilibrium headspace  analysis.   The percent recovery of
tK ft(JUeo°UtldS was "Is*161*  in aqueous-phase  wastes than organic-phase wastes.
I   v4n0Us wastes, the relative partitioning of a particular compound into
^ge   * Phase depended  on that  compound's  solubility in water.  Therefore,
o^Us °!Jnt8 of nonpolar compounds were present in the vapor phase when
^&ruc Phase wastes were tested.  Conversely,  the light mineral  oil in the
   the     e wastes appeared  to  sufficiently depress the vapor pressure of
      compounds so little partitioning was observed, even at 90 °C.

  *tne  data obtained with the GP&T analysis was generally lower than that
        1*-0 the steam distillation for some of the waste types.  However,
      ofrecoverles °f the compounds  were usually caused by the  incomplete
   6 t»»o  Pheno1 and naphthalene from the OOP.   In fact, only 30 to 50* of
        compounds were effectively removed from the OOP.
1,  *|VL
c0* t>teviETp*T unfortunately did  not follow  any of the trends observed in
k  Ut»da ^8 three methods.   It gave both high  and low recovery of the
^ *he com   botl1 aqueous- and organic-phase wastes.   Trends were observed
O e«ch   und recoveries with the ETP&T.   The  highest recoveries obtained
Si  ene Waste tyPe were usually the  two most  volatile compounds,
Hoi^W    lori
-------
                          Precision of Test Methods

                                                                     i nti^
      Precision  estimates  of the methods were based on the percent reJB
 standard  deviation (%RSD)  of the percent recoveries,  as calculated fr<)
 triplicate  analysis  of  the six waste forms.   The 3.RSD values of all n  ,{t
 compounds were  then  averaged together for each waste  type.  These data
 shown in  Figure 2.

      Note that  the average *RSD values of the nine compounds were not
 dependent on  the waste  form tested.   However, the averaged precision  (
 dependent on  the test method chosen.   For example, the %RSD values °*.-j(
 ETP&T and equilibrium headspace analyses were generally higher than tn
 of  the steam  distillation  and GP&T analyses.

 Conclusions

      Comparing  the percent recoveries and precision for each Candida*
 method made it  clear that  each method had advantages  and disadvantaff6 '.^
 The advantages  of  the equilibrium  headspace  analysis  were simplicity'
 turnover  rate,  and commercial  availability of automated systems.   T°e
 advantages also applied to the ETP&T.   The steam distillation method  ^
 usually had the highest recoveries of all the components In all wast^ ji
 tested, as well  as being precise.   The GP&T  had high  recoveries of nt°
 the compounds in all waste forms tested and  good precision.

      The  disadvantages observed for  the equilibrium headspace analV*
 included  low  recoveries of the nine  organic  compounds from all the
 synthetic waste  forms (especially  the organic phase wastes)  and im~ 0 ^
 precision.  The  ETP&T had  erratic  recoveries  of the compounds and a aas
 not precise.  This may have  been caused by the fact that the method
 never optimized  for  use with concentrated wastes such as the six
wastes.   In the  steam distillation method, the test and  sample a   y
 were  very labor  intensive.   Further,  that system could not be eas  »s
cheaply automated.  Finally,  the DOP  used to  suspend  the synthetic *
 for the GP&T analysis wastes is a  suspected  carcinogen.
                                                                .   gp&'
      Ideally,  one candidate  test method will  accomplish all  of tne * t|
 However,  one method may not  accomplish all of the purposes for whic fl
test method was  developed.   Therefore,  the choice of  any test metho
depend on the situation.   For  instance,  the  data obtained with the  $
equilibrium headspace analysis  might  relate more accurately  to ""
measured  from storage tanks  or  other  enclosed areas than would
obtained  with the GP&T or  steam distillation.
                                                                    X
     When a final test method  or methods are  chosen,  future  develop flfy
will focus on optimizing and automating volatile organics analyse*- ^J
analyses  will be measures  of total organic carbon and total  organ*  ^t
removed from the synthetic waste rather than  labor-intensive sPeci
 like those described here.
                                   410
-------
    CONCENTRATION OF EACH COMPONENT IN THE SYNTHETIC  WASTE  TYPES
                                         Concentration
type                                           (Ug/g)
   Phase  aqueous

'60U8 sludge

 Ute aqueous

     sludge

  ic

                                                 3,000

                                                   650


                                                     15


                                                 2,000

                                                 2,000

                                                 8.000
                             411
-------
I
K
  100 n





   eo -


   7O -


   60 -


   50 -


   40 -


   .TO -


   20 -


   10


    0
                    1
                        / \
^
 S
 v
 S
                           1$
                           A$
                            s
                                       K^
                                 1
                                                      s
                                                     / 5
II
                                                     /. S
                                                           1
                                        ^
                                        fa
      DILUTE      MIXED-PHASE  AQUEOUS SLUDGEORGANIC SLUDGE


C7"71  HEADS PACE
                                                               ORGANIC
                               _       WASTE TYPE  	
                               rs^l  ETP&T         W7\  STEAM
              Figure  1.   The average  percent  recovery  of all compound5'
      100 -





       80 -


       70 -


       60 -


       60 -


       40 -


       JO -


       20 -


       10
        O J
        >7\
        S
          DILUTE
          HEACSPACE
                 MIXED-PHASE  AQUEOUS SLUDGEORGANIC SLUDGE    ORGANIC
                                   WASTE TYPE
                                 ETP*T
                                                        STEAM
                   Figure 2.   The average %RSD recovery of  all  comp°
                                       412
-------
 R MISSIONS  FROM  HAZARDOUS
WASTE STABILIZATION
         Percin
             Control Division
D,St p0us Waste Engineering Research  Laboratory
Clnl,tnvironmental Protection  Agency
  Clnn*ti, Ohio
&n<|

Or  ,

     n|jental Engineering Division
         oration
       Triangle Park, North  Carolina


   ili  v?1atile Organic Compound  (VOC) emissions  from  the  laboratory
    g 2ati.on of surrogate hazardous  wastes and from the stabilized wastes
     t?Uri r>g have been determined  to be significant.    Two  types  of
    d ^c hazardous wastes were tested: 1} an  inorganic  "waste" made  of
       Water, and 2) an organic "waste" made of soil,  water  and  latex
       dge.  The "wastes" were doped  (spiked) with acetone and trichloro-
       'TCE)» representing volatile compounds, or n-pentanol  (amyl
d ed wit rePresenting a semi volatile compound. These wastes were  stabi-
 ^l,  th a mixture of 50%  fly ash and  50%  portland cement or  lime  kiln

J"1ssion measurements found that stabilization does  not  reduce the VOC
hi  es  !hs^9nificantly.  In fact, during the processing  (mixing) of the
M^6r'th   emiss'ions from tne wastes being stabilized were  significantly
eih^°ut ttn froni tlie same material being treated in  the same manner but
th  S1ons   addit1on of solidification agents, e.g.,  blanks.  The VOC
V ^sto Were Pr°portional to the amount of the constituent present in
  *ater  • but t(ie vapor pressure of the constituent and  its solubility
            weakly affected emissions.
                                 413
-------
Introduction

     The 1984 Resource Conservation and  Recovery  Act  (RCRA)  amendment^
require the promulgation of regulations  limiting  the  air ^"l^Lnts
RCRA facilities and operations.   At the  same time,  the  KLKA  amenanen   $
also'  equed that no liquids be disposed  in landfills  and specific -JS
be prohibited from land disposal.  The "no liquids" and "waste  banning
regSlat ons are creating an impetus toward stabilization as  a waste t ^
Tent.  Unfortunately, hazardous waste stabilization has also been  ld_en
fied as a possible air pollution emission  source  to be  regulated.   i   „
determine the significance of the air emissions from waste  stabilizati^
VOC measurements were performed during laboratory-scale stabilizei   dy
a mass balance performed.  The laboratory research allowed  for  the si  „
of several  variables that could affect the total  emissions  and  the en
rates.

Experimental Design

     The  study was designed  to determine the VOC emissions  during the
mixing and  curing of the waste.  This required two different measure** v
systems   a  wind ?unnel  and a headspace  sampling  system   The wind t  ^
tests were  conducted by  putting  a weighed quantity of waste ino a T
gallon  pail,  and  then  adding water,  dopant  (spike) and stabi izer i   d
pail.  The  pail was  placed in a  Lundberg mixer and  mounted in the w   ^
tunnel.   The mixer was turned on,  and mixing continued for  sixty to   -
minutes  while  the emissions  were measured.  After mixing  the mixing   #
was  removed from  the pail, the  pail  removed from the wind tunnel, an
pail  sealed with  the lid.
      After mixing,  the pail  was placed  in  the  sample  venting system,^
 which pulled air through  the pails  headspace at  a  controlled rate.    y
 emissions from each sample were measured for three to four  hours, &
 three or four days  for 30 days after  stabilization.

 Equipment Design                                                     J

      The wind tunnel system consists  of a Lundberg mixer which was ^
 into an eight by eighteen (height x width) inch  rectangular duct.    ^
 lop plate of the duct was tightly fitted around  the mixer head to § ,
 ai? infiltration.  A hole was cut directly below the mixer  head, an
 clamping mechanism was constructed to hold a five-gallon pail  tign
 against the duct and mixer.  A hinged door was installed next  to t,
 m?xer to allow  addition  of materials while the system was in  opera  ^
 One end of the  duct was  open to the air and the other end had  a ra    *
 vented the emissions  outside the building.  The duct was heated to
 condensation.

      Wind tunnel gas  flow and  VOC  concentrations
  ously.  The  air  velocity was determined by measuring the Pres
  acros   a  calibrated  orifice plate.  The VOC concentration was
  with an in-line  gas  chromatograph with a flame lonizat ion d etecto   ^
  FID)  in accordance with  USEPA  method  18.  Measurement of the VUts
  made  after the orifice plate to ensure good mixing.                   .5

       During  curing  and while waiting  to be  sampled,  each waste  (Pa  $f>
  connected to the sample  venting system.  Ai r was ^^""V^llV'
  the pail, i.e.,  over the stabilized waste,  at  approximately  one
  minute.  This venting was  to simulate the expected VOC  loss  from
  waste during curing - maybe from  the  disposal  site.
                                     414
-------
Co   Every three or four days for three or four hours, each waste was
Kepnectecl to the head  space sampling system.   Emissions from each waste
  * measured by the same GC/FID used to measure the emissions during the
 Xln9 {dual  column).

 Xperimenta1  Approach

to d The research program was a parametric design which varied five factors
1)  eterri>ine their effect on the VOC emissions during waste stabilization;
     *? characteristics, 2) stabilization agents, 3) constituent (dopant)
       ^'  4)  vapor pressure and 5) constituent concentration.   A statis-
      was consulted to help design the test  matrix.

^s a    tyPes  °f wastes were prepared for this study.  The first waste
the s  inor9anic waste consisting of soil  (25%) and water (75%);  while
(6,2«?Cond was  an organic waste made of soil  (18.8%), latex paint solids
4 Wa ' and water (75%).  This was to determine if nonvolatile organics in
  aste would affect the VOC mobility.

of fj  ° stabilization agents were used to treat the wastes.  Equal  amounts
On   y a$h and  portland cement, or fly ash and lime kiln dust were added
     ^ua1  wei9'1t basis to the synthetic  wastes.  To assure satisfactory
    ''zation would be achieved, preliminary  test were conducted  with
c^or   ee  Or9anic  compounds  were selected for this study;  acetone,  tri-
to r °ethylene  (TCE)  and  n-pentanol  (amyl  alcohol).  Acetone  was  selected
VaPor    ent  a  P°lar  voc» so1uble in water and highly volatile.   The
to reJDressure  °f acetone is  greatly affected  by  water.   TCE  was  chosen
*4$ sLresent  a  water  insoluble VOC.   For a semi volatile  organic,  n-pentanol
to be lected  because  it is  nontoxic  and  has  a  high  enough vapor pressure
     Pleasured  by the same  system used  for TCE and  acetone.
            the concentrations  of  the  three  organics  were  varied  to
    ft       emissions  were  related  to waste concentrations.  Three concen-
    °ns were used;  1,  2  and  4% by weight.

     hus»  the  test  matrix was the following:

        2     Wastes     (organic and  inorganic)
        2     Stabilizers  (cement or  lime  kiln dust, and  fly  ash)
        3     Chemicals      (acetone,  TCE, and  n-pentanol)
        3     Concentrations     (1, 2, 4%).

   9 wi S  and  sometimes  triplicates were performed  on  each  combination,
      'th  blanks and mass balance checks.

 ^n, -^ this  large test matrix  and the many  data points  for  each  combina-
 "^ roll normous data  set was collected and stored  on  computer.  The
P     a|istic  way to interpret this data was  by  statistical tests.
 Vln,
    1lnental  Results

Of" 1n ?hlaboratory  results determined that a  significant  portion of  the
d! the w   wastes are emitted into the air during the mixing and curing
up*nta!,a1ste'   Table 1  shows  the  percent loss  for acetone, TCE and
P» stl%lri  durin9 the mixing  and  for the 30 days  afterward {curing  period).
*r Cent      noted tnat the percent  VOC  during the curing  period is the
         the  remaining  VOC.  The blank is the emissions  from the  spiked
            the addition of the stabilizer.
                                 415
-------
     The type of waste, i.e., organic and inorganic, does affect the V
emissions.  During mixing, emissions from the organic waste were less
than the inorganic.  During curing, however, organic and inorganic waste
emissions were similar.
     The act of stabilization had its greatest impact on the VOC
during mixing of all the components.  During mixing with lime kiln
ash, emissions were far greater than emissions from the blanks, i.e.*. ,
mixed without addition of stabilizers.  Emissions during the mixing »itt
the portland cement/fly ash did not differ significantly when compared ,
the blanks.  There is a greater than 90% probability that the method uS
for stabilization does not affect the VOC emissions from the treated
waste.  During curing, the acetone, TCE and n-pentanol  emissions shove"
no significant difference.

     Vapor pressure is a very significant factor during mixing, but n°   (
during the curing of the waste.  TCE emissions were 30% greater than t ,$
n-pentanol during mixing, but less than 5% while curing.  TCE and
have greatly different vapor pressures, but are both relatively

     Solubility may have a great effect on the VOC emissions.  When c° j
paring TCE and Acetone a significant difference in emissions is obsefV
Unfortunately, it cannot be determined if the difference is due to s°'
ity or vapor pressure.  The partial  pressure of acetone drops quickly
when mixed with water.
     There is no clear trend to whether VOC concentration in the    e
results in a proportional  emission rate.   The results seem to indicat
this, but there were differences which could not be correlated over 9
of test data.  Generally,  the emissions from 2% VOC waste were twice
emissions from 1% waste, and half the emissions from 4% waste.

Conclusions
                                                                     a
     The emissions from hazardous waste stabilization are significfn
this treatment process cannot be expected to control  the VOC emission
before stabilization is complete.
                                   416
-------
                       TABLE 1

              MEAN PERCENT OF VOC LOST


             % During Mixing         % After Mixing*
Binder
fement/Fly ash
Li|"e/Fly ash

Acet
83
68
58
TCE
60
50
30
Pentanol
68
12
8
Acet TCE Pentanol
40 28 22

sr mixing is given as percent of VOC remaining in waste
 after the mixing process is completed.
                        417
-------
INTEGRATING SAMPLER FOR
HAZARDOUS POLLUTANTS IN LIQUID STREAMS
R. G. Merrill, Jr.
J. B. Homolya
D. H. Doerle, and
L. W. Cooper
Radian Corporation
P.O. Box 13000
Research Triangle Park, NC 27709

F. E. Butler
Quality Assurance Division/Source Branch
Environmental Monitoring Systems Laboratory
Research Triangle Park, NC 27711

ABSTRACT

     Under the Resource Conservation and Recovery Act (RCRA) nonpo*
sources such as ponds, land treatment areas and wastewater treatmen
systems are the focus of research activities designed to assess
volatile organic compound (VOC) emission characteristics.  Current!/
regulatory activities require industrial discharge limits on a mass
per year basis.  Analysis of the volatile organic compounds in the   .
process streams entering these ponds and treatment systems may be   j,
as an indicator of the potential for hazardous VOC emissions from s
systems.

     Multiple grab samples must be taken to represent VOC in a stre
with changing flow or organic component concentration.  When the
process is extremely cyclic, grab sampling and analysis of a suf*
ciently large number of samples to represent the stream become lal)
intensive and cost prohibitive.

     Automatic samplers solve many of the inadequacies of grab
ing techniques.  Automatic samplers allow integrated samples to
collected without increasing labor costs.  However, no available
sampler provides automatic adjustment for flow variation and inte'
gration of volatile organic compounds from liquid process streams-
                                   418
-------
   ^n integrated Liquid Sampling System (ILSS) was developed  to meet
   6l?uirements of VOC sampling of liquid streams.  Laboratory
   ation of the ILSS was successfully completed.  Field evaluation
 e  r£°nned at a municipal sewage waste treatment plant.  Results of
    tests are reported in this paper.
       Environmental Protection Agency (EPA) is involved  in pro-
        emissi°ns standards to control several emission sources of
 So  e organic compounds (VOC) listed as hazardous wastes. Under  the
Pon(j r°e Conservation and Recovery Act (RCRA) nonpoint sources  such as
EOCU ' lflnd treatment areas and wastewater treatment systems are the
t«tj °f research activities designed to assess VOC emissions charac-
to £pA   •   Air emissions from waste holding ponds are also a concern
-------
in a gas tight sample reservoir with no headspace.  There are
currently no commercially available automated samplers that provide
zero headspace and time integrated sampling for volatile organic
compounds in liquid process streams.

     An Integrated Liquid Sampling System (ILSS) has been developed by
Radian Corporation to meet the requirements of VOC sampling of liquid
streams.  Laboratory and field evaluation of the ILSS was successfully
completed. The ILSS sampler consists of a sampling probe to draw
sample from a source stream such as an inlet weir of the treatment
plant and a syringe-driven sampling head connected to this probe.
syringe drive consists of three (3) 25 mL syringes mounted on a
housing containing a stepper motor used to control the rate the
syringes withdraw sample from the probe (Figure 1).  The ILSS has the
capability to sample at a constant rate and therefore requires a
constant flow rate source stream for accurate sample acquisition.
FIELD EVALUATION

     A field evaluation of the ILSS was performed at a municipal
sewage waste treatment plant.  Clogging of the ILSS syringe tips by
suspended solids was minimized by placing a coarse screen around tn
sample probe inlet.  Clogging was further reduced by adjusting the
flow rate in the probe bypass leg of the sampler. The majority of th*
flow and solid matter was forced through rotameter-#l (Figure 1)•
slower flow rate in the bypass leg of the sampler carried mostly    .
aqueous sample through a coarse screen, through rotameter-#2,  and P*
the ILSS syringe sampling tips. Waste stream input to the treatment
plant was monitored during sampling and found to range between 9.3
9.5 million gallons a day through each of two weirs.  The ILSS pr°be
was placed at the confluence of the two input streams to the waste
treatment plant (Figure 2).  The ILSS samples were collected at a
constant flow through the sampling probe and a constant sampling ra
setting on the ILSS syringe driver.

     Initial field shakedown and mechanical performance tests wefe
executed during three days of sampling.  The sampling probe and
syringes were flushed with source sample before sampling was sta*ce
Grab and integrated samples were taken simultaneously. Field coll®
ion of grab samples was done every 30 minutes over the sampling
period.  For equipment shakedown, every fourth sample was analyz®

     Standard VGA analyses using EPA Methods 601  and 602  were
performed on samples and appropriate laboratory blanks collected
during the equipment shakedown.  Analysis was performed using a &
chromatograph with a photoionization detector (GC/PID) and a gafi  .
chromatograph with a hall detector (GC/Hall) for both detection *n
quantification.
                                   420
-------
jj  Several mechanical improvements in the ILSS were  identified  over
nCTC°Urse of the field sampling shakedown.  These  improvements


   0  Conversion of the slipstream-to-syringe connections by
      replacing narrow bore hypodermic needles with  316  stainless
      steel leak-tight fittings and wider bore tubing,

   0  Collection of grab samples from the outlet  of  the  ILSS probe
      overflow rather than from the source stream confluence,

   0  Analyses of each grab sample in duplicate;  and

   0  Analyses of well mixed syringe samples in duplicate.

6ve     improvements were instituted during a 4th  day of  sampling.
ay  6en (17) grab samples were collected in duplicate on the fourth
    Three ILSS syringe samples were taken simultaneously.   One  grab
        eack duplicate pair and each of the ILSS  samples was
       by GC/PID for compounds listed in EPA method  601  and 602. All
   6a were analyzed in duplicate,


   Ts AND CONCLUSIONS

   B ata were evaluated to determine if concentrations of  target
    n«s differ in simultaneous samples taken with  the Integrated
     Sampling System (ILSS) and grab methods.  The  results  of  the
    ^Ple analysis were mathematically integrated  over  time using
  *ia  matneniatical models.  All of these calculation methods gave
  en i results.  Average results for the 4th day grab samples are
    ltl Table 1.  The average of the results from all three  ILSS
  |sses analyzed in duplicate are given in Table 2.  Higher  differ-
    ^ere seen between the two sampling methods for  compounds
       in samples close to the quantification limit of  the  GC
       n«thod.   The results of the grab and ILSS  samples were
"u to ,°n a time scale to examine the concentration-time  dependence
p^cted  ntify any Periodic, linear, or non-linear  trends.   Plots  for
«*5Ur6   Con»pounds from this series of tests are  shown  in
ry the      6.  The variation in concentration of the target  analytes
Vigfegwastewater stream with time was much greater  than anticipated.
*te me    8rab samPle concentrations from 10 to  200 ng/L  per hour
     asured in extreme cases.

>0|^0Ur»dd agreement (less than 10 % difference) was  found  for most
l^latl S SamPled with grab and ILSS techniques.   Even  with the wide
v^een n in concentration over short periods of  time the  difference
h PetCfi    ILSS and grab sampling methods was never larger than
^S S}j nt>  The results of laboratory and field  evaluations  of the
      w that the ILSS can acquire integrated liquid samples
                                 421
-------
containing volatile organic compounds.  The comparison of ILSS and
weighted grab samples shows that the two sampling techniques are the
same within experimental error.

     The next step in the development of an integrating liquid
sampling system for hazardous and process streams is the inclusion of
a flow sensing feedback loop to provide automated control of the
sampling rate.  Such a sampling system will provide true automated
proportional integrating liquid sampling capability.
                            REFERENCE

1.   Federal Register. 49, 43261-80,  1984.
DISCLAIMER
     This paper has been reviewed in accordance with the U.S.
Environmental Protection Agency's paper review and administrative
review policies and approved for presentation and publication.
                                  422
-------
Ji,
ro
CO
                   Hotameter  #1


                        Manual
                        Control
                        Valve
                       Coarse
                       Screen
                        Manual
                        Control
                        Valve
                      Pump
                                     Filtei
                                                      \
Rotameter  #2
                  Wier  1
                                                                 Motorized
                                                                 Syringe
                                                                                                        Sampler Location
                                       Flow
I    no   I
NFlowmet/rs
                                t

                                                                                                Flow
                                          /
             t
                                                                                                                 Flow
                                                                                                     Grit  Chamber
                                      Plant
                                     Influent
                                                           Wier  Z
                          Figure  1. Field  Probe and ILSS Configuration
                                                                                Figure Z,  Sampling location  at Farringlon Hoad  Treatment Plant
-------
                                             e
                                             o
                                             o
                                                   zz
                                                   zo
                                                   II
                                                   1C
                                                        	 AVB CBAB
                                                        •  QHAB glPLJGATI 1
                                                        •  GRAB  HIPUCATI-Z
                                                        — - BTB AVB
                        Bunpl*
Figure  3.   ILSS  and  grub  flata for
\
M>
9
                                                                       (3
                                                                       e
                                                                       u
     3D
     SB
     26-
     34
     K
     EO-
     1B-
     15-
     14-
     12
     10-
      B-
      6-
      4
                                                                                 	 AVE  CHAB
                                                                                 +  CHAB HEPLICATt-1
                                                                                 «  GHAB REPLICATE-Z
                                                                                 - - AVE  SYRINGE
                                                                                          (CC1T  )
 Figure
                    111          It         «         B>
                       Sample Urns
             ELSS  and grab  data for  perohloroethyLene
NJ
   £
   s
   a
                                                e
                                                o
                                                u
                                                        	 AVI QBAB
                                                         •  GRAB  KEFIJCATE 1
                                                         •  GRAB  REFUCAE-B
                                                        -- STR AVE
<
w»
9
13
O
     34

     3O
     SB
     BS
     Z4
     ZZ
     so
     IB
     15
     14
     IE
	 AVE GHAB
 *  GHAB  HEPLICATE-1
 4  GHAB  HE PLICATE-E
- - AVE SYHDJGE
                                                                                                                                        Ul
                                                                                                                                     1V.SS «.n
                                                                                                                                                           Tor CTiVorotorm
-------
                                                                                         ZLSS FIELD EVAUSIAITOH
                                                                                     AVERAGE GRAS SAMPLE RESULTS
ro
en
                               Compound
                                                      C-l
                                                             G-2
                                                                     G-3
                                                                            G-4
                                                                                   G-5
                                                                                          G-6
                                                                                                 G-7
                                                                                                        G-8
                                                                                                               G-9
                                                                                                                      G-10
                                                                                                                              G-ll   G-12
                                                                                                                                             G-14
                                                                                                                                                     G-15
Freon (CC1 F)
3
Chloroform
1 . 1 , 1-Ir Lchloroe thane
Perchloroethylene*

7.5

13.
3. A
1,0

s.6

12.
2.1
2.1

7.7

12.
3.1
1,5

8.3

13.
3.6
3.9

8.4

13.
3.3
1.9

8.5

16.
3.0
3.9

7.6

20.
4.0
3.0

7,5

17.
5.2
A.O

8.3

21.
5.2
6.16

10.

31.
7.
10.6

9.0

33.
6.8
11.

7.9

24.
16.
5.9

3.7

27.
8.3
12.

9.13

32.
5.5
9.2

                              Bo rtportable data were acquired for Grab-13
                              Only one replicate was reportable for Grab-9
                              * Fhotoionlzatlon. detector results
                                               TABLE 2
                                 PRECISION OF SYRINGE SAMPLER RESULTS
                                                 DAY 4


Coupound
Fceoti 
-------
VALIDATION OF ANALYTICAL METHODS FOR DETERMINING TOTAL
CHLORINE IN USED OILS AND OIL FUELS
A. Gaskill, Jr., E. D. Estes,
D. L. Hardison and L. E. Myers
Research Triangle Institute
Research Triangle Park, North Carolina  27709

and

8. Lesnik
U.S. Environmental Protection Agency
Office of Solid Waste
Washington, DC  20460
     A current EPA regulation prohibits the sale for burning in
nonindustrial boilers of used oils and oil fuels contaminated ab°v%gjj'
specified levels with certain metals and total chlorine (the "Burn   f
Regulation),  When burned as fuel in a small boiler, the contami*1811
be emitted  to the ambient air at hazardous levels.  This regulati011 .$
establishes a rebuttable presumption that used oil containing more  * J
1,000 ppm total chlorine has been mixed with halogenated solvents   ^
hazardous waste.  One means of establishing the chlorine content °
oil is to test it for total chlorine.

     To provide regulatory agencies and the regulated community tfl M,
accurate, cost effective methods for determining chlorine in used    t
EPA/OSW has evaluated several analytical techniques, prepared wfi
methods for the most promising and subjected these to a collabor
study to generate precision and bias statements for each method
methods evaluated were based on microcoulometric titration, X
fluorescence spectrometry, two test kits, and five analytical
oxygen bomb combustion.

     The results of this collaborative study are reported here.
this study, final versions of the test methods will be proposed
inclusion in the ASTM Book of Standards and the Federal Register
acceptable  for compliance with the "Burn Ban" Regulation.
                                   426
-------
CHLJ?*TION  OF  ANALYTICAL METHODS FOR DETERMINING TOTAL
   RINE  IN U
        IN  USED OILS AND OIL FUELS


l.  T
   •"•ntroduction
the y!??re  than  1  billion gallons of used lubricating oil are produced in
^U      States annually.   A significant fraction is sold for burning in
tyPica??nindU8trial  re8idential. commercial,  or institutional boilers,
°     y  after blendin  w
    a              re8idential.  commercial,  or institutional boilers,
°ils f y  after  blending with virgin nos.  4 or 6 fuel oils.  These used
t=0ntanjequently  arrive  at fuel reprocessing or blending facilities
^68e ;nated with  chlorinated solvents,  lead,  cadmium, and arsenic.1
been   °ntaminants may  be Present as a result  of the oil's use or may have
6*CePti    throu8h mixing with hazardous waste.  Reprocessing, with the
     ion  of rerefining,  typically fails to remove these contaminants.
        burned  as  fuel  in a small boiler,  the contaminants or their
   g  t   byProducts mfiy be emitted to the  ambient air in quantities high
 11 Ud IP present Potential hazards to exposed individuals.   This concern
C0tlt«min    t0 Promulgate a regulation prohibiting the sale  for burning of
  cifj  ted used oil-2   The first  person to claim used oil  fuel meets the
  por»-  *ion levels must obtain  an analysis or other information to
     c the claim.
             rule  (40  CFR  Parts  261  and 266)  establishes a rebuttable
ii- nal    that U8ed oil  containing >1,000  ppm total chlorine is mixed
i ecau8fe°8enated hazardous waste and,  therefore,  is  a hazardous waste.
tHb°rato t0tal hal°gen content  cannot  easily be determined in the field or
.''e x.. ry» and because  chlorine is almost always  the major contributor to
          content, EPA has agreed to interpret  "total  halogen" in the
         as total chlorine.)3

?'«.,nuadditlon' the rule establishes a  specification for used fuel
M0"! a]^ ed °il not mixed with hazardous  waste)  that  is  essentially exempt
ti*k aii0re^ulati°n and may be burned  in  nonindustrial boilers,  provided
   cine TA    levels for designated toxic elements,  flash point and total
       14,000 ppm) are not exceeded.

4 *°tcema £0n*equence of this regulation,  the regulated  community and the
   ni fu  autnorities will need to determine  total  chlorine  in used oils
     meet     The ideal analytical technique  for this  determination
    *te g  8everal performance criteria.  The  technique should  be
    lQ00Uch that it can determine if the chlorine content  is above  or
      L- %*!&'   Ifc should also be precise  such that  replicate  analyses
      "t h 6d 8amPle yield reproducible  results.  It should be  rapid,
'U  ^« T     and inexpensive.  The used oil-to-fuel  industry is  somewhat
V ^Pme quirin8 in many cases onsite decisions as  to  acceptability of
^n»i'tl ^e tl'   Ideally«  the analytical determinations should be  carried
»JUyt   *ield  to quickly identify contaminated oil and oil fuel.
\ ^e Var46 testing procedures should be simple enough  so that users over
W6 c*n b6ty  °^ 8ki^  and experience levels can perform the analyses.
   CatOty I exPected to include generators,  burners, truck drivers,
         technicians,  and enforcement personnel.
S *o D
^*4te(   Vide regulatory agencies and the regulated community with
\^ ha8°8t  effective  methods for determining chlorine in used oil,
      for ®^aluated several analytical techniques,  prepared written test
          the most promising,and  subjected these to a collaborative

                                 427
-------
study to validate each method. ^i5   The methods evaluated were based on
microcoulometric titration, X-ray  fluorescence spectrometry, two  test
kits, and five analytical  finishes  to oxygen bomb combustion.

     The results of the collaborative study are reported here.  More ^
70 laboratories representing generators, collectors, recyclers, rere-
finers, burners, testing laboratories, and regulatory agencies partic-
ipated.  The study was a cooperative effort with ASTM Committee D2 on  ^
Petroleum Products and Lubricants.  Based on this study, final versi011
the test method will be proposed for inclusion in the ASTM Book of    .(
Standards and the Federal  Register  as acceptable for compliance with *
"Burn Ban" Regulation.

2.  Approach

                         2.1  Methods Evaluated

     The instrumental microcoulometry (MCT) method  involves an
technique in which an oil  sample is first burned in a combustion
under oxygen to convert the chlorine to chloride.   The chloride is
reacted with silver ions in a titrating cell, where the silver ions 8
replaced coulometrically.  The total current required to replace  the „,
silver ions is a measure of the chloride present in the injected  samP

     The X-ray fluorescence (XRF)  spectrometry method involves a
nondestructive technique in which  the sample is placed in a dispos** ^
plastic cup covered with an X-ray  permeable membrane and irradiated
X-ray radiation from a source within the instrument.  The irradiated  (j
sample will then emit fluorescent  element specific  X-rays whose inte $t
can be quantified and correlated with the element's concentration i°
sample.
     The bomb combustion method  is based on ASTM Method D808-81.
sample is pressure oxidized by combustion  in a  stainless  steel
containing oxygen.  The liberated chloride is absorbed in a  sodiun1   j
carbonate /bicarbonate  solution contained in the bomb.  The ASTM ^et^e
uses a gravimetric precipitation of chlorine as silver chloride as
analytical finish.  However, this finish was found to lack adequate
sensitivity at the 1000 ftglg level.  As a  result of the pilot st**Z' &&
several alternative analytical finishes for chloride were evaluate*  f
included in the new method.  These are mercurimetric titration, B* .,
nitrate titration, ion-specific  electrode measurement (ISE), a fe*
cyanide colorimetric measurement, and ion chromatography  (1C).  *  ^e
measurement principles for each  of these finishes are described in
test method.
                                                                     *°
     The field-portable disposable test kit method is applicable
such kits.   Both kits operate by first dehalogenating chlorine f
chlorinated organics (solvents,  polychlorinated biphenyls fPCBs]
reacting the oil sample with a sodium naphthenate mixture in *       *
catalyst.  The chloride released by this reaction is then extrac
an aqueous buffer and titrated to the mercuric nitrate endpoint.

     In the Dexsil Corporation kit, called Clor-D-Tect 1000, the
sample and reagents is fixed such that a dark blue color  results
chlorine content is below 1,000 /*g/g and a yellow color results
concentration exceeds 1,000 /lg/g.  There is a relatively  narrow
zone over which the color change occurs.

                                   428
-------
       Chemetrics  Inc.  kit,  called Quanti-Chlor,  involves  largely the
        stry  as  the  Dexsil kit with the  exception that the extracted
        s  titrated into an evacuated ampoule  containing mercuric  nitrate
    °n to  a yellow or  colorless  endpoint.   The  chlorine concentration in
      over a  range of  750 to 7,000 /lg/g  is  read from a scale  on the
                          2.2   Study  Design
   TK
     e  collaborative  study was  designed to satisfy the requirements of
                on  Petroleum Products for determination of method
        and bias-   Tnis  required  that a minimum of five laboratories
         Per method  (or method variation as  in the case of the  bomb
  10  S  and tne  test kits).   For this  number of  laboratories,  a minimum
     amPles was  required.  Each laboratory was  also required to  analyze
         in duplicate.  This would result in a minimum of 100 determi-
     Per method.

          2.3   Selection  and Preparation of  Test  Samples
   Th
     6  10 °H samples consisted of used crankcase  oil, used metalworking
        No- 6  fuel oil  blended 9:1 with used crankcase.  These  oil
        believed to be representative of the  types of oils or fuels that
     *   1? require  testing under  the  EPA burn ban  regulation. Each
    a.atcn was  spiked with a mixture  of volatile,  semivolatile,
V  ^ed     Snd  inor8anic chlorine.  Background  chlorine levels were
f6*6 incl  8everal techniques, including neutron activator analysis, and
 Ot e*ch      in  the  exPected  concentrations.  The expected concentrations
       °il type  are given  below:

   ^                                        Concentration, flg/g
Npl
V^OrtM                              A8°        92°     1,527   3,045
   °il K  8                            268      1.871    29,540
H      blend                          320      1,498     3,029

I o c*llyCentrations  are  intended to cover the range of levels of chlorine
    Ppto 8een  in  uaed oils<   The regulatory presumption threshold,
       '  i8 bracketed by the range selected.
    te  ^a°oratory  received approximately 30 mL of each sample.   This
    8 v   n enough  for the required determinations.   The numbering of
    be  *  rai*doinized with respect to concentration so that no trend
       evident.
          «nd  Discussion
 1  1)
 ^ij^otl'8 Were  received for 58 of 77  samples sets shipped to 66
 ^^tri   °nly  two  laboratorie8 reported results for the bomb-
 Vct6(|   ^ method,  one with order of magnitude differences from the
   0(>Vaa    8*   °ue to the  small data  set,  no further evaluation of this
         attempted.
      Ult
 \  ^d i,   *0r  the metalworking samples  were 60-100 percent lower than
    6*tl  Sl1 methods.   The  effect of the water matrix of these samples
            factor in  the  low results.   For the bomb combustion methods
            step was  probably incomplete,  resulting in little or no
                                 429
-------
r
determinable chloride in the combustate.  The water matrix absorbed
fluorescent X-rays of chlorine resulting in low results for the XRf
method.  There was no obvious reason for the low results by MCT
system is capable of boiling off the water prior to oxidizing the
organics.  Water consumed the sodium reagent in the test kits before
dehalogenation reaction could occur, resulting in low values for tb*
analytical steps.  For these reasons the metalworking sample
not included in the estimation of precision and bias for these
review of earlier research  suggests that a water level of 25 percent
be tolerated.  The final versions of the methods make note of this *
as provide for use of spike recoveries to monitor for this effect.

     For the fixed endpoint test kit 95 percent of all results  (53/5
8 laboratories) were correctly reported as < or >1,000 /ig/g.  Th
no within laboratory disagreements and only a few between laborato

     Data for the other methods were evaluated using a SAS program   t
first identifies and removes outliers, stabilizes variance and  calc
precision and bias for each method.  Both the repeatability  (sing^e
operator or within laboratory standard deviation) and reproducibil*
(repeatability plus between laboratory standard deviation) were
calculated.
                                                                   tilt
     The precision and bias information for these methods are sun*1*  c}
in Table I across all concentrations tested.  In general, the rePrlfl8 (
bilities (Z RSDg's) are 2 to 3 times greater than the repeatabili'
RSE^s) as expected.  The results for the MCT and XRF methods shov
negative biases and excellent precision within and between labors'
The bomb finishes show somewhat worse performances.  The relative*/ ^
interlaboratory precision of the 1C method could not be explained-
ISE method has a high bias which also could not be explained.
four laboratories submitted results.  The silver nitrate and mercy
nitrate titration methods showed similar performance except  for
<1,000 figlg where the silver nitrate method bias is quite high, °"0
-------
6.

I,
    ences
*ranklin Associates,  Ltd.   "Composition  and Management  of Used  Oil
          in  the United  States."   Prepared  for  the U.S.  EPA/OSW,
Un<*er Contracts 68-02-3173  and  68-01-6467.   1984.

 •s- Environmental  Protection Agency.   Federal  Register.  50,  49164-
      November 29,  1985.
        A.,  RTI  to  D.  Friedman,  EPA/OSW,  personal  communication,
        29,  1986.
        A., E.  D.  Estes,  D.  L.  Hardison.   Evaluation  of  Techniques
 °r Determining  Chlorine  in Used Oils,  Volumes  I-IV.   Prepared for
198 U'S' EPA/OSW» under Contract No.  68-01-7075, WA No. 58,  June
o**' A'' E' D' Estes' D<  L-  Hardison  and L.  E. Myers.  Validation
j  Methods  for Determining Chlorine  in Used Oils and  Oil Fuels.
  epared for the U.S.  EPA/OSW under  Contract No.  68-01-7075, WA  No.
  • June 1988.
       TABLE I.  SUMMARY OF PRECISION AND BIAS INFORMATION
            FOR CRANKCASE AND FUEL OIL BLEND SAMPLES
        Nitrate
                 Labs
                    10
                      7

                      6
                      4
                      5
                      5
                                 Z  Bias
  -513
  -814

 24132
 86164
 14143
  9119

-11118
 412
 615

2119
1517
 816
 814
                          I RSDB
 1516
 1114

43119
36116
24110
21110

 1417
                              431
-------
EVALUATION OF PERFORMANCE OF CARBON MONOXIDE
MONITORS ON HAZARDOUS WASTE INCINERATORS
Larry Edwards and John Randall
Radian Corporation
Austin, Texas  78720-1088

R. Rollins/ T.J. Logan, and M.R. Midgett
Environmental Monitoring Systems Laboratory
Research Triangle Park, North Carolina  27711


     To gather background documentation on the performance character1
carbon monoxide (CO) continuous emission monitors at hazardous waste
ators, three instruments were evaluated at three different
they underwent performance certification testing, several instrument
problems meeting the drift criteria.  However, the principal problem
because of the monitors and test procedures, but because of the mo
operation of the incinerators.  When the incinerators were burning    -
waste, the CO concentrations remained well below 10 ppm; the CO was   ^
at the quantitation limit.  When solid waste parcels were burned, C
trations spiked to 5000 ppm and above, and the monitors operating °n
0-100-ppm or 0-500-ppm range were inadequate.

     The EPA Reference Method 10A, currently applicable only at Pet^,j $
refineries, was used for the relative accuracy testing for CO.  Me*    "
worked well when the integrated bag concentration of CO was less tn
but at higher levels, the CO analysis showed greater variability » *n
method limit of 1800 ppm was sometimes exceeded.  Long-term operatic
strated the need for adequate environmental protection for the in8
faithful removal of water to prevent the formation of aqueous HCl *
instruments and pumps, and frequent attention {at least once per w
knowledgeable technician.
                                    432
-------
             EVALUATION OF PERFORMANCE OF CARBON MONOXIDE
               MONITORS ON HAZARDOUS WASTE INCINERATORS
           regulations1 specify that temperature, pressure, oxygen, and carbon
    *    °'  shall be used as indicators of destruction efficiency in hazardous
        nerators (HWIs).  Most commonly, these parameters are measured with
         m°nitors,  but although the technology for the first three variables
      establ*8hed,  the performance and reliability criteria for CO continuous
       m°nitors (CEMs) at HWIs have not been documented.  The purpose of this
      °n was to gather background information to help establish that
    ntation.
         C0 CEMs were evaluated at three different HWIs.  The CEMs were
           undergo an initial performance certification test, to be operated
          weeks to establish longer term performance characteristics, and to
              performance certification test to identify any deterioration.
         e schedule could not be followed at any of the three sites due to a
t 8%  Ihf problems which were believed to be typical of CO CEM operation at
 °est*bli  type °^ information addresses the primary objective of this study,
       ieh performance and reliability criteria for CO CEMs at HWIs.

   BniC01np0nent of the performance certification testing involved
     r n<3 tn«  relative accuracy of the instruments by comparing the instru-
     - *dings  to the concentrations measured by an EPA reference method.
       f\f*
     e  ° Promulgated reference method for CO applicable to HWIs.  However,
I ^Ittafc1106 Method 10A is a non-instrumental method for determining the CO
4 ** r«£i   n °f a 3°-min integrated sample of flue gas that applies to petro-
 "^Ustfai.6*^68*2  Therefore, a secondary objective of this study was to
N      te the  transferability of this method to HWIs.
 *Cj4
       011 °f the Incinerators

          ee Bites visited were designated A, B, and C.  Site A was a process
          that burned liquid waste, but could also incinerate process off-
            l  gas was fired as the primary fuel to maintain operating
      } th %g*' temPerature J •  Tne liquid waste consisted of several types of
   *t a      process gases were most often formaldehyde.  The stack gas usually
           la* oxygen.   Except for periods of maintenance, it was operated 24
     t\jr   Tnere was a secondary combustion chamber but no scrubber.  Stack
        es were typically between 700 and 800«C.
      feg
     ati  Snd c were  commercial  HWIs  capable of solid and liquid waste
     Pri°n*  Eacn  nad  a  fixed hearth primary combustion chamber and used oil
     »a  ry f«el.   Liquid wastes could also be aspirated into the chamber.
           WSre ducted  to a  secondary combustion chamber, which was co-fired
     h Wa  9*8'  site  B  used  tne  neat  from the Process to produce steam, part
          inJ«cted  into the  secondary off-gas for particulate and acid-gas
          e C Jlad * ^*y limestone scrubber followed by a baghouse.   Both
  i     tj^  an^ ^ were  operated  more  or less continuously,   in each,  the
           duced  a  container (usually a cardboard box or fiberboard drum) of
          Directly into  the fixed hearth chamber, where the temperature was
                                   433
-------
normally between 800 and 1000°C.   The container and waste smouldered for 30-K
sec before the vapors ignited.  Solid ash was periodically  removed from the
chamber.  Often  the  operator injected a new container of solid waste into the
incinerator  only when certain parameters (e.g.,   off-gas  CO  or SO?  or HC1) wert
below  predetermined levels.   Liquid waste  was normally sprayed into the chanbe
only when the parameters were at acceptable levels.

Experimental Methods

     Four different  extractive  CO monitors supplied by EPA were used at one
time or  another  during this  study.  They all used non-dispersive infrared
(NDIR)  spectroscopy  for  analysis,  but they were not all identically configure.
For example, the Beckman ijodel  866 compared the 1R signal  from two cells — oa
being  the sample gas and the other being sample gas after it had passed throng:
a catalytic  reaction cell where all of the CO was  oxidized to CO?.  The Dasibi
model  3003 also  used two cells, but one beam was  optically "scrubbed"  (that it,
it first  passed  through  a cell  of CO  gas) before analysis.  The  Anarad  model
500R compared the sampled gas to a reference cell.  The Infrared Industries
model  760 chops  the  beam and compares the analytical  signal to that from a
full-scale reference CO  cell.  As will  be  seen  later,  CO?  is  a slight
interferent  in NDIR  instruments,  and  because of their different designs, each
monitor  responded differently to the GQ^ interference.

     The  gas extraction,  conditioning, and delivery system  was built around a
TECO model 600 gas conditioner.   A sample of the stack gas was extracted throif
a stainless  steel  probe  and  introduced  into  the conditioner.   The  TECO  condi-
tioner chilled the sampled gas  to remove the water.  The dry gas was then
through a Teflon*-coated pump and routed into a glass distribution manifold.
The manifold was maintained at atmospheric pressure by having excess gas vente:
to the air through a rotameter.   Each of the instruments used its  own pumping
system to draw a portion of  the gas from the manifold and, after analysis,
vented it to the  atmosphere.   A high-pressure blow-back  system was activated
every 30  min to  ensure  that  the probe and sample line remained unobstructed.

     The  electrical  output from each  instrument was routed  to a  multipen strip
chart recorder to  provide a  permanent and complete performance record for eac!
monitor.  The signals were also sent  to  a data  logger, where  the printout and
storage of data  could be controlled by  keyboard commands.   A  data  point was
taken every  2 or 5 sec and stored.  Five-minute and 1-hr averages  could be
calculated and either printed out  or  stored  or  both.   Generally, the 5-min
averages  were printed out and maintained during the certification testing.
During the longer  performance periods,  only the hourly averages were retainei

     The  instruments were calibrated  and challenged with standardized gases
prepared  and certified by a  reputable vendor.  Because the instruments  were tc
be operated on the 0-100-ppm scale, zero-level,  50  ppm,  and 90 ppm CO concen-
trations  in nitrogen were used  for calibration.  The  gases  were  introduced vis
a three-way  valve  at the exit end of  the probe.

     The  EPA Method  IDA  sampling and  analysis was conducted as specified in tb
proposed  regulations.2  Briefly, dual sampling  trains  pulled  the dried  flue
gas from  the distribution manifold, bubbled  it  through alkaline  potassium
permanganate to  remove acid  gases, passed it through  a Teflon*-coated pump ani
through a rotameter  for  an estimate of flow, and finally injected  it into an
evacuated Tedlar® bag.  The C02  concentration in the manifold was  also  measure:
                                     434
-------
  B   yrite* kit.  Later, the gas  in  the bag was  analyzed as called for in
  y    d through a colorimetric  reaction with  para-sulfaminobenzoic acid
       with a bench-top spectrophotometer.
       certification procedures followed  the  intent  of  similar certification
   irur®s  Published  by  the EPA for S02/NOX CEMs.3   These procedures include
   and*0*1 error  tests, two-hour zero  and  span drift  tests,  twenty-four hour
       span  drift tests,  instrumental response time tests and relative
       tests.

          Discussion
tlls Ub °re  ttlese  instruments were transported to the field,  they were tested in
*jcutac ratory.  Performance  certification testing,  except  for the relative
   An    6St*  Was carried out on three instruments  (Anarad,  Dasibi, Beckman).
        monitor« which had a range of 1000 ppm.  did  not  meet the drift tests
        Tb-e laboratory  testing demonstrated  that all three  instruments were
   ef*ect S  C°2  interference; the  greater the  percentage of C02.  the greater
        '
              increase was  rapid  at  low concentrations of C02.  but nearly
             C02.  With  about  10% C02  and  nitrogen in the instrument,  the
     Blo      read about -4 ppm,  the Beckman monitor about +2 ppm,  and the
 Otlitor Wglt0r ab°ut +25 ppm.   Later,  in  the  field,  the Infrared Industries
       as found to read  about +7 ppm for similar conditions.

C" l° Ppte ^ where only  liquid waste was fired, the  CO was generally less
u"*8'  Th     A Performance certification was carried out on all three instru-
*M0st no^-o0^61^1  results  are summarized in Table  I.   Because there was
^ 9Uve »     n th-e flue gas,  CO was injected into the sampling line for the
      t C^Uracy  testing to establish  a  mid-range CO concentration.   Response
            instruments were about 1  minute.   The  Dasibi and Beckman monitors
             but the drift  in  the Anarad instrument,  originally noted in the
       ac8tudy-  continued to be a  problem.   In fact,  the readings  during the
  Culatem *U,  ai      accuracy  testing with artificial injection of CO appeared to
S    fr0m    Ough th-e drift  problem with the Anarad monitor disqualified those
        g  any  furtl»er consideration.   The  reading of  the Dasibi and Beckman
        B agreed satisfactorily with  the Method IDA results; therefore, if the
         aate< Working as  designed and the concentration of CO is mid-range,
    ?f HWl Ppears to be reasonable  for determining  CO concentrations on this
      °ie'0fTh<:  reProducibility  of  the dual bag samples was good and differed
      og w   °ine  cases  by  more  than +2.5%.  And  in  that  one  case  where the
             i3'8%. one  determination of the triplicate bag gas concentration
               ,from  the concentrations determined by the other two analyses.
            an irregularity in the analysis of this bag.

           th*
        t oge  L results at sites B and  C  were similar to one another,  they are
          uet   "  At these sites, the Anarad monitor was replaced with an
                8  ^  fflonitor'  At bottl sites,  the Dasibi monitor passed the
           o            and  response-time parts of the certification.   The
        eite r Bailed the four drift tests at site B and one (the 24-hr zero
             '  The Infrared instrument  failed  two drift tests at site B
                                   435
-------
(the 24-hr zero drift  and  2-hr calibration drift)  and one at  site C (the 24-hr
zero drift).   At site C,  the Infrared monitor failed  the  mid-range  calibration
test (but passed the high-range  test).  All  response  times were 70 sec or  less.
Thus, the Beckman and Infrared instruments experienced some difficulty  in
meeting the proposed performance standards,  but  the Dasibi instrument met all
criteria easily.  The relative accuracy test at  the two  sites  presented prob-
lems of a different  sort,  and  no valid  relative  accuracy results were obtainei

     At sites B and C, a box or drum  of hazardous  waste,  if available,  was
injected into the incinerator once all monitoring parameters were well  below a
predetermined level or once the previous parcel was completely consumed.  The
parameter that  most  commonly controlled the introduction of the solid parcels
into the incinerator was  the CO  concentration  in the  flue gas.   When a  parcel
of solid waste  was  introduced  into the  primary chamber,  the CO concentration on
the readout of the operators control board rose very  rapidly and usually went
off scale.   It stayed off scale for a few seconds  to  a few minutes.  When it
fell back to about 10 ppm and another parcel was  injected,  the CO  rapidly rose
again.   Liquid waste was  sprayed into the  chamber either separately or  when the
CO reading was  low, but  it  usually had  only  a  small effect on the flue gas CO.

     Thus,  the pattern of CO concentrations  was  one of very low readings
(below 10 ppm) punctuated with spikes going well off  the  0-100-ppm  scale on
which the instruments  were set (see Figure 1).   Readings  from  other, noncerti-
fied instruments and calculations based on the concentrations  measured  in
Method 10A bags indicate  that  these peaks  frequently  reached  concentrations of
5000 ppm and probably much  higher.  The CO monitors used  in this study  had a
voltage-limited output, and when the range was exceeded,  the  voltage  and its
indicated proportional CO concentration remained fixed at  that  maximum level.
Those levels were between 250  and 500 ppm  for all  three  instruments used at
sites B and C.  Thus, if the data logger received  a signal such as 400  ppm and
calculated an average  for the period, which  was  often between 100 and 200 pptn,
it was clearly invalid.  When compared to the 30-min  integrated bag sample
taken during the Method  10A testing for relative accuracy,  and when there had
been a CO excursion well off scale, the measured CO concentrations  in the bags
always exceeded the  integrated readings of either instrument,  as they should
have.  The  relative average readings of  the instruments over the 30-min
intervals simply  reflected  the maximum voltages put out by each instrument.
Thus, except for periods when no solid wastes were being  injected,  the  relative
accuracy test could not be  carried  out.   When no solids were  incinerated, the
CO levels remained below 5  ppm,  and instrumental averages when corrected for
C02 were within +2  ppm of Method IDA results.

     Several other results were noted.  In an attempt to  overcome  the CO spiking
problem, a surge  tank  was placed in the sampling line to truncate the very high
spikes and still allow for  accurate time averaging.   But  to keep the CO concen-
tration within  the bounds  of the instrument,  the surge tank had to be  so large
that the response time exceeded the 15-min requirement.   When  the  high  CO spiiei
occurred during the Method  10A sampling,  the 3%  precision between the dual bag
determinations demonstrated at petroleum refineries was  not consistently
achieved.   Also, the precision expected for  the  triplicate analysis of  each bag
was not always  met for CO concentrations of  200  ppm and  greater.  The stated
limit of the  method,  1800 ppm, was exceeded 20% of the time in this study at
sites B and C.
                                      436
-------
Ua    e
 i   X
          * at site C, a total hydrocarbon  (THC) analyzer was also  installed
   °Perated with the CO monitors.  In general, THCs remained below 5 ppm  and
      showed little or no variation while the CO concentration was  spiking
a c  off scale.  On occasions  (perhaps 10% of the time) a THC response  tracked
 dfl  SPike-  THC spikes, usually lasting from 30 sec to 5 min, also occurred
 he ^endently of any CO spike.  Maximum THC concentrations sometimes  exceeded
    *librated scale range of  200 ppm.
           conclusions may he drawn from this demonstration testing of CO
  la, t8 at hazardous waste incinerators.  Some of the conclusions refer to
    8 not previously discussed in this brief presentation.

       Operating methods used at many solid HWIs preclude the conventional
     C° CEMs<  EPA is considering requiring either two monitors or a  dual-
           with a low range of 0-200 ppm and a high range of  0-1000 ppm.  The
  re 8 aeen at these sites indicate a high range of at least  10,000 ppm would
          to cap the CO spikes.

       The CO CEMs appear to be certifiable if the instruments are in good
        c°ndition and if the process operates at a stable CO  level during the
     e *ccuracy test.  The most commonly failed test in this  study, other
     l*tive accuracy, was the 24-hr zero drift test.
   «
   t   The CO CEMs and the data acquisition system must be well protected
    emP«rature fluctuations (a maximum of +_ 5°C is suggested) and dust (fly
     f°Wer fluctuations and outages were also common, and for long-terra
       9' the system must be able to withstand such incidents.  Therefore,
  * qu9eable desi9n and installation is imperative and preventive maintenance
     *Hfied person is required at least weekly.
._  «
«°t8tUj.  The sample conditioning system must be reliable and must never allow
.*1e g4* to condense in the instruments.  When condensation occurs, HCl in the
 ° •ign81*bsorbs into the droplets resulting in corrosion which normally leads
      1 instability and ultimately to total instrument failure.
^  «
^6*t8  When fluctuations in CO concentration are not extreme, EPA Method 10A
jV, v.  ° be satisfactory for relative accuracy determinations at HWIs.  How-
*  I8ioen V6ry hi^h co BPlkes occur during the 30-min period,  the accuracy and
   *v*dn °f the m«thod demonstrated at petroleum refineries was not always
                                  437
-------
References

1. U.S. Environmental Protection Agency, 40 CFR 264,  Subpart  0  (264.347}-

2. U.S. Environmental  Protection Agency,  Method 10A — Determination of
Monoxide Emissions  in Certifying Continuous Emission  Monitoring Sys'temfi
Petroleum Refineries, Federal  Register,  52(158). 30675-30681, August  !?•

3. U.S.  Environmental Protection Agency, "Performance Certification Te0(
EPA 625-79-005, Continuous Air Pollution Source Monitoring Systems  as up**
in the Federal Register, Vol. 48,  No. 102, page 23608, May 25,  1983.

4. U.S. Environmental  Protection Agency,  40 CFR 60,  Appendix B,  Perform811
Specification 2.

Disclaimer

     This paper has been reviewed in accordance with  the U.S. Environ»e11 ^
Protection Agency's  peer review and  administrative review policies  and a'
for presentation and publication.
                                    438
-------
        I.  Results of Performance Certification Testing at Site A
2
  Cr
  of
Beckman Dasibi Anarad Performancel
'""eter (Units) 866 3003 Specification
lt Zero Drift (%)
lr Calibration Drift (%)
lr Zero Drift (%)
lr Calibration Drift (%)
186 Time (Minutes)
^=^!uracy (%)
0.440
0.658
0.220
0.279
0.67
5.46
0.100
0.799
0.008
0.232
0.67
11.1
43.1
35.7
21.8
4.4
0.75
	
<2% of Span
<2.5% of Span
<2% of Span
<2% of Span
<15 Minutes
<20% of Mean
lteria in
Specification 2 used as guideline because of lack
         criteria for carbon monoxide CEMs.
                                 439
-------
 Figure  la.    Typical  CO concentration pattern;  0-100 ppm range
              (5  hour  period,  time moving to the left)
     J/l/M
Figure lb.
Typical CO concentration pattern; 0-500 PPra
(16 hour period, time moving to the left)
                                            rang*
                                  440
-------
                                                     EMISSIONS
    ce
 St. B !   Blackley,  and


                 Park,  North  Carolina
 s::j.  -,
           'gett
                                                     '
                  'rotection Agency
                  Branch (MD-77A)
                Park, North Carolina
w0rk
    encyCfEPA? ^ W" fUnded by the U'S'  Environmental
    •«! .  ^A)/nd Was  de^gned to formulate and evaluate a metl
     tweC  r ?g emlSSlons  fro™ *e  incineration of hazardous
  Other   BT   metals'   In the futu«.  this  method may be used
                            gr°ups to measure
                           PrVate gr°ups to measure toxic »«als
                  gWtraL°rwiJhate? ^ "" baS6d °n the US6 °f a «««
                               nltrlC  ""Aydrogen peroxide,  water,  and
                     manganate  used as the absorbing solutions.
                        was evaluated by  spiking  the  absorbing solutions
                        t, and digesting these samples  either with
^ptton   in
-------
microwave  pressure-relief-vessel digestion was approximately 20 times faster
than  the conventional  Parr®  Bomb digestion.   Digestion of spiked municipal
solid waste  (MSW)  baghouse fly ash samples by microwave techniques gave
recoveries of 70  to 100  percent for all metals except beryllium which
averaged about 54  percent.
                                       fe
      For all  microwave digestions the  best results were obtained by heating
the samples for a  total  of about 15 min in 1-to 3-min power cycles.

      Based on the  results of the experiments  described above a method
was recommended and has  subsequently been tested in a field study.  During
this  field study  the metals  thallium,  silver,  antimony, and barium were
added so that a total  of 16  metals were evaluated.

Introduction

      The U.S.  Environmental  Protection Agency (EPA)  is currently investigate
toxic metal emissions  from stationary  sources  as potentially hazardous air
pollutants.   The metals  of concern in  this study are lead,  zinc,  phosphorus,
chromium,  copper,  nickel,  manganese, cadmium,  selenium, arsenic,  mercury,
and beryllium.  If EPA makes a determination  to regulate emissions of some
or all of  these metals,  appropriate methods of sampling and analysis must be
available  to  accurately  quantify the emissions of the metals in stack gases
from  stationary sources.

     The Environmental Monitoring Systems  Laboratory (EMSL)  of EPA in
Research Triangle  Park,  North Carolina,  is developing a method for sampling
and analysis  of toxic  metals emissions.   This  report presents the results of
the laboratory study of  the  toxic metals  emissions measurement method.  The
objectives of this  study were as follows:

           Determine  the  applicability  of  a Modified  EPA Method 5  (MM5)  train
           with nitric  acid/hydrogen peroxide,  water,  and acidified potassiuii
           permanganate as the absorbing solutions for the collection and
           measurement  of toxic  metals.

           Determine  the  best methods for  digesting samples  collected in
           the  proposed train.

           Determine  the  precision and  accuracy of the proposed laboratory
           analytical techniques.   The  techniques consisted  of sample
           preparation  followed  by analysis by  graphite furnace atomic
           absorption spectrometry (AAS) or ICPES.

     The method formulation  and testing were conducted in several stages.
The initial effort  focused on defining appropriate sampling  and analytical
procedures.    Then a laboratory study  was  conducted  using the proposed
methods.  After the  initial  laboratory studies  were  completed,  a  field test
was undertaken  to determine  how well the method works in the field.   The
results of the  field test will  be  presented in another paper by Osmond et.
al in this symposium.
                                   442
-------
  *periaental Methods
Th
    Snra"l4ng and analytical  methods  evaluated in this  field and

           ofeIamPliP°Sed  after  S  th°rough  lit^ature  search.    The
            o          i                                          .
 >a'ed accordin  ^ ^ ^Ple PreParatl°n.  and analysis were  then



      ng with rnq^s      ?
-------
     The  conventional heating  digestions of  the  filters were perform^
 Parr9  Bombs.   Each  spiked  filter was placed  in a bomb, and 4 ml of
 concentrated hydrofluoric  acid and 4 ml of concentrated nitric acid "e j
 added.  The bomb was closed  and heated in an oven  for 6 h at 150 C.  •* p
 cooling to room temperature, the sample was  filtered and diluted to *"
 with de ionized water.

                                                                      ^
     The  microwave  heating digestions were done  in microwave pressut* ^
 vessels (PRVs) which were  purchased from GEM Corporation.  Spiked fi * . (<
 were placed in_the  PRVs with 4 ml of concentrated  hydrofluoric acid
 of concentrated nitric acid.   The PRVs were  closed and microwaved
 5 min  at  100 percent power.
     In addition, filters with 0.5-g samples of fly ash added were
by the microwave method described above.  Both spiked and unspiked
were prepared.  The method followed was as described above with the
exception that longer microwave times were required for complete °*i£fl
A typical heating program was a total of 9 to 15 min on 100 percent P
with the digestion done in intervals of approximately 2 min.  If
heating Is used, the PRVs will vent due to the high pressure with
possible loss of sample.
     All samples were analyzed using AAS and ICPES or Neutron
Analysis (NAA) .  Graphite furnace AAS was used for the analysis of   y
levels of arsenic, selenium, and lead; and ICPES was used for the fl
of all other metals.  NAA was used to confirm ICPES and AAS resul
ICPES analyses were performed by following EPA Method 200.7 (40 CFR
136, Appendix C) .  High levels of iron and aluminum were present in
the samples .  The samples were diluted so that the iron and alum*n
were less than 50 ppm to make their interferences generally negl*S . p
the samples containing iron and aluminum are not diluted to belt>w
interfering element correction factors should be applied.

                                                                  of*
     A retention experiment was designed to determine the abilicX  ^ gf
absorbing solutions to retain the metals.  This experiment consis      (
spiking the metals of interest into impingers containing the va*i   ijj '
absorbing solutions and pulling ambient air through the impinge*      j
rate of approximately 10 1/min for a period of 2 h.  The nitric     e
-------

of Z6" iT  heatinf di*estion °f the impinger contents,


           S        "  ^  ° 104'6
         o

   !al1 recovery of9f 8S SrSV"? ^ <° 104'6 ^"^ With
 3 6 Stl°n ranged frd** ?T ?no  1ReCOVeries for the microwave heating

 a "7 Percent g? f^M M  i        PerC6nt With a mean overa11 recovery of

 iT^y-U of variance   Si C°mPa^Sr °5 **' tW° meth°ds Was  do»e by «•
 t 6T b°th due to S!  i    I anft,1ysls of variance showed a significant
       action S™ K ^lementuanaly"d and the digestion method used, but

       the percent «      theS6,!:r effects Was »ot significant,  ihus,
      ^thods  the ""Y"7  Was/ifferent f". element to element and

    the «vera     I   J   """    the """ Erection.   A t test showed


                                  s:
                             ^
     '-r ^f- -sr^^«^1^" s^
   ^tho^^ Recovery of 94.6 percent.  A statistical comparison of the
         showed no  significant effect due to the  digestion method used
          ath.               Sh SampleS in &en&ral showed lowe^ mean
  Ltles of *?£ "P^ed ^lte" Containing no fly ash.  However, average

  L * f°t all  T      n ?5 Percent we" obtained with the spiked fly ash

       .   A oL ?M tS eX?6Pt berylliun!-  ^ ™.n beryllium^ecovery was

       may form^   T'K?  "ati0n f°r the l°W beryH^ recoveries is that
        «y form insoluble  complexes that precipitate out of  the sample
                          indicated that  the absorbing solutions


           centarMr,met:alS U?der simulated samPllnS conditions.
        The S?        i   Percent)  showed greater than 90 percent

        <= acid b^ W6re ClCTd ty thorouShly ri^ing them 3 times with
       carrv   be^ween experiments.  Most metals showed less than


              rW    SrPleS'  A fCW We" Sllghtly higher'  but o
                        5 percent  carryover.
 ' oywas^               '  lt: Was found that the  ^trume

 **Vi!Uiveri snfuf         the 10SS  of mercury f«m the solutions.  That
     hln + ,oPlkine and Analysis  day, the mean mercury analytical values
        i 10
             percent of the true value.

V  blfis

  6
                  . results  di""ssed above  for each of the  elements of
                e by request from the author.
                              445
-------
 Conclusions  and Recommendations
                                                                      MI
      Several conclusions  and  recommendations have been made  regarding  j
 use  of  the proposed  sampling  train,  digestion  methods,  and analysis i"6
 These are given below.

      1.   Both  conventional heating  and microwave digestion  methods g*
          recoveries  of 100 + 20 percent when  digesting combined nitr* ^
          acid/peroxide and water  absorbing solutions.   The  microwave
          was faster, however, and can be used to speed the  digesti°fl
          process.
                                                                     ti$
      2.   Both  the Parr®  Bomb and  microwave pressure  relief  vessel I"6 ^
          of digesting spiked filters gave recoveries of 100 +  20 PetCg
          However, the microwave digestion method is  20 times fastet
          this  is the preferred method when the equipment is availfl^
                                                                     t fl"
      3.   The retention experiments  that were  performed indicated tn» rt«''
          significant loss of metals is expected from the trains ui
          sampling conditions.

     4.   The acidified potassium  permanganate solution retained tn
          that was spiked into it.

Disclaimer
                                                                   o
     This paper has been  reviewed  in accordance with  the U.S. Envif
Protection Agencies peer  review and administrative review
approved for presentation and publication.
                                   446
-------
       S   FROM HAZARDOUS  WASTE INCINERATORS
       OR   OPERATING   PROBLEMS
        Santoleri
          INC-
       Meet ing, PA
Qln9 ha  e_source Conservation and Recovery  Act  (RCRA)  cove-
Pr rated       waste  incinerators  regulates  the  emissions
*t  Qrilv  K  :he  incineration  system.  These  emissions  are
1]  burneri    acid gases generated when halogenated compounds
fluids   fS Wel1 as the Parfciculates generated  from burning
ill  s-  M*    dges' or solids containing  ash or  inert mate-
t&  9UedC    lncinerators  in operation  today  may have been
%°btain     ially without air pollution controls,  in  order
*te' lr>  m      part  "B"  Permit'  emission  standards  must  be
to^ t0 m  y cases air  Pollution  control systems have been
C0h! °PerT    RCRA standards-  (1)   Liquid  Injection Incinera-
r^  tol  si       steady   state  and  allow   the  air  pollution
   S *^d       design  to  be based  on combustion gas flow
     nq  0mP°sitions generated  in the  incinerator.  Systems
     rat^    cycling  conditions without properly  designed
     s i   ai^ P°llution  control  systems have created major
^tGe(3 by    maintaining good  emission control.  These were
CC frOm ?u Physical nature of  the  feed,  the heat  release
W0k01s f     wastes  fed  into  the  incinerator,  or draft
V" fea S5 air  fl°w.  Most solid waste incinerators are
    rWr/   e sYstems  operate under  a negative pressure  in
  a it:Y     -hamber.   With batch  feeding,  the  cycle  ,  the
  % 4litv   * batch' the heat available  per batch,  and the
  •.?*«  in     :he wasfce determine the  instantaneous heat
       t     incinerator.  This  also establishes the air
       thL°f the  system-  Waste  types and degree of turbu-
       to tuPr   ary  charr|ber  are  critical to  the carryover
       hi  K  APC s^stem- Hign velocities created by opera-
        if dl^aft  conditions also  create particulate car-
        onmS not contemplated in the initial design. These
           Bating problems will  be the subjects covered in
                          447
-------
 Introduction

      In  any waste  incineration system*  whether  hazard^"
 under RCRA or  TSCA*  or non-hazardous under state incinera6
 standards*  the details  covering waste  quantities/ phys*C]
 characteristics  and  other pertinent data  covered  in Table
 are  needed.  (2)

                          TABLE 1

                        Waste Data

                   Chemical composition
                   Heat of combustion
                   Physical data (if not liquid)
                   Viscosity
                   Corrosivity
                   Reactivity
                   Polymerization
                   Ash/Inerts content
                   Fusion temperature of ash
                   Combustion product analysis
                   Nitrogen composition
                                                          $
 This  aids  in establishing the  necessary  design  details'jg/
 well  as  operation of the feed  system/  temperature  cont*  jj
 air  flow control  and scrubber  controls.   Additional  dfje ^
 information related to the total system as covered in Ta
 are necessary.
                          TABLE 2.

             Design Details of the Incinerator

                    Physical dimensions
                    Materials of construction
                    Feed device
                    Injection method
                    Auxiliary fuel system
                    Combustion air system
                    Waste Heat Recovery System
                    Quench system
                    Scrubber system
                    Controls/ Monitors & Alarms
                                                    .t
     With the detailed waste  data  and flow  rates/  .l
be possible to determine the  capability of the in
system.  Many systems initially designed for one set °  ^v
stream conditions but with process plant changes creaj
waste streams require a closer  review of the  capacity
incineration system to  handle these new wastes.
include waste flow rates/  composition and heating va
installed combustion air system  determines the ovef~
release  capacity of  the system as well  as  the v°
                             448
-------
the  st*on  products into the  downstream scrubber system.  If
  1"  a^e higher  chloride  concentrations  in  the waste »  the
      of acid  gases generated increases affecting  the design
      P°.llution  control  system. The  physical  data  for  the
    . solids will  affect  not  only  the  feed  system  (ram  or
      but  also the existing  pollution  control system.

      and Operating  Problems

    "eview the air flow  rates  and  methods  for determining
by gT rates.  In  induced draft fan  systems/ this may be done
    9ck  flow  measurements  at the  discharge  of the  induced
     •    or by tne fan amperage.   In the  forced draft  sys-
    air  flow can be  determined  with flow meters in  the  air
      lines/ pressure drops  across  burner windboxes or motor
      e-   Measurements should  be  made  of  these air  or  gas
      anc^  tests  should be conducted to determine  that these
     matcn those established  by the   heat  and  material
       based on  total heat   inputs     and  excess   air  rates
      ' • On many  induced draft systems/  air in-leakage down-
      °f the afterburner  has created difficulty in relating
    n98 flow  measurement with combustion  gas velocity  in
 ions,  fln?rator chamber. With oxygen readings at both loca-
      this  air  in-leakage can easily be obtained.

     !?*;  following problem areas related  to  emissions   are
    q      have occurred  in  various  incinerator  systems  ope-
  riiiq   Under interim  status. These problems  were discovered
      normal  operation prior to or during the test  burns.
        listed below:
             Particulate  Loading  (Fig.  1)

        High Velocity  in  Primary

             Design  - Minimum  Particulate  Size  or
                       Density  not   considered   in
                       determining  flow area.
                     - Primary/Secondary  Areas undersize
                     - Poor  Seal  Design ( Induced Draft
                                          System)
                     - Burner Design &  Location

             Operation
                     - Excessive  Draft  Control
                     - High  Excess  Air  Rate
                     - Loading  Rates &  Cycle of
                       Loading  cause High Draft
                       Operation.
                           449
-------
     B.   Bottom A§ji Removal

               Design  - Dropout Volume Undersize
                       - Velocities in Primary &
                         Secondary High
                       - Separation Zone Inadequate

     C.   Secondary Chamber

               Design  - Low Velocity
                       - Horizontal Configuration
                       - Dead Zones (Low Velocity
                         Regions)
                       - No Provision for Ash Removal

     D.   Ash Build-Up

               Combustion Chambers
                       - Horizontal Design
                       - Freezing or Slagging Ash (Figure

               Waste Heat Boilers (Figure 3)
                       - Flame Impingement
                       - Insufficient Ash Cooling
                       - Tube Spacing

2.   Scrubber System (3)
     A.  Design
                         Venturi not  designed for  l°a<
                         particulate size.               _.fltl>e
                         Particulate sizing data assumed
                         than established by testing.
                         Control for turndown not p^
                         Improper material selections
                         bag  collector  or  electrostati
                         precipitator.
                         Quench Design                   ^
                         Location of Particulate Removal
     B.   Operation
                         Poor maintenance of Instruments
                         Pressure Drop Controls.
                         Recycle stream pH control not
                              450
-------
       Combustion Efficiency

        Design
   B
        Operation
                      Primary Chamber undersized
                      Low L/D in kiln
                      Low solids residence time
                      Burner System Type & Location
                      Location and Design of Atomizers
                      Feed System
                      Cycles for loading extended
                      High instantaneous loading
                      Poor air control (Forced Draft or
                      Induced Draft)
                      Maintenance of Atomizing Nozzles
        Thermocouples

                   - Design/ Location/ Calibration

              ng Train

                   - Design r Maintenance
^
                   - Design/ Calibration/ Maintenance

              Sampling (4)

                   - Stack  Design (Sample Port Location)
                     Cyclonic Flow
                   - Mass Balance Check
t(j  It i o
ly *0|Jito  llnPortant that a sampling procedure be established
\L*t ae?r.emissions of acid gases and particulates especial-
bft* ar*  *s contained in the waste. If chlorinated hydrocar-
       also contained  in the waste  streams/  a  means should
         -d for measuring the  overall efficiency of  the
         Jn other  words/  the  HC1 entering and  leaving  the
        should be  sampled so that  the efficiency of  the
        :a/i be  determined. This should be done if the  expe-
       sion rate  is  above  the  four  (4)  Ibs/hr.  Absorber
                may  be  modified  to improve efficiency.
 *<*i
                           451
-------
     Prior  to formal  test  burn, observation of  the ^
       An interim status hazardous waste  incinerator  pla J,
a test burn  to determine  the adequacy of the  system  to
RCRA standards. Figure  4  shows a typical arrangement
primary and  secondary  stages. Figure  5  shows the  o
conditions  that  normally  exist  in  this  incinerator
Note that  the primary  stage operates in  a  reducing
phere;  i.e., there is insufficient air to completely  ox*  tj
the organic  wastes.  The main  feature of  this design   * f 01
reduce the turbulence in this chamber to prevent  carryoy   «
ash particulates. Many   units of this design are  operati ^
industrial and   hospital  service without air  pollution j^
trols.  The secondary chamber has air introduced at the   ti'
port",  the  exit  of  the primary.  At  this point,  the  vo  M
lized gases  from  the primary chamber mix with oxygen
air and combustion of these  volatile  gases  occurs.  Au
heat is available with  the addition of a burner in the
dary chamber  at  a location  close to  the  exit of the   t
port".  This  insures good mixing  and  flame  surface
this point.  (5)
       The incinerator used a ram-feeding system to
wastes  in  40 gallon  fibrepacks  and carton  boxes. A
sludge was also injected through atomizers  into this F jj J
chamber. Due to the  high viscosity of the sludge  as  ^tj^
ash contained in  the  sludge,  no  attempt was made  toare0*
this with  a  valve.  The sludge was pumped with  a Pr?jjcin*
cavity pump  from  the  storage tank directly  to  the *?„<#*
tor. This  resulted in a constant heat load from the »* JrJt
10 MM Btu/hr. The solid wastes were fed at  15  minj}
vals. The  average load  was  approximately 250  - 300
heating  value of the  waste averaged  6000  Btu/lb-
this would vary between  5000  and 8000  Btu/lb.  The
of the   solid wastes   also varied considerably depe     f,
the source and the composition of the wastes.        00*
      Due  to the  length  of  time  between  feeds  of t^
and the  variation  in heat content  and volatility
difficult  to maintain control in the system. Tne
installation was  supplied with only  the incinerator
and stack. Forced  air was supplied via the  unae
blower, the flame port blower and the combustion avc t
the burners  in both  chambers. When the  owner  deci°  ^
the system to burn industrial wastes,  including »
                             452
-------
   es'  it  was  necessary to provide a scrubber for HC1  remo-
    Since  this  was  a  starved  air  design*  it  was decided that
       ate en>issi°ns  would not be a problem. The  specifica-
     were based  on a  system capable of  incinerating  at  a
pt     state of 20  MM Btu/hr. A  waste heat  boiler was  also
ift y* ded for  heat  recovery  and also serve  as  a primary quench
to p   Scrubber train. A low-energy venturi  quench was  used
Xij^.?0-1-  the gases  exiting the boiler at 500 deg. F to appro-
The  elv 190  deg.  F entering  the  HC1 packed tower scrubber.
rieq Astern as  it  exists today  is shown  on Figure  6. The
Pteve e Pressure  (draft) necessary in  the  primary  chamber  to
the i   fugitive emissions  during loading is  maintained  by
^       d draft  fan.  The  inlet damper of  the fan  is contro-
       the pressure controller located in the primary  cham-
        the rooa-ifications were made with  the boiler and air
  tQ  on  control  system* the  stack was converted  to  an
 n thncY vent  stack.  This was to open  only  if the pressure
loss e Primary went  positive*  the  i.d.  fan  failed or due to a
8e"tie !..  water flow  to  the  boiler  or scrubber system.  At the
8t°P  »  e that  the vent stack  opened f  all waste feeds  would
^ to K ever any waste remaining  in the primary would conti-
5*teiv ?u rn since  the residence time for solids was approxi-
J>t .X 30 minutes.  The initial modifications did not provide
S°0ts £  seals at the loading areas. Each time the  guillotine
 CcU  pened  for the ram feeder* a sudden  inrush of air  would
        to the draft.
       Figure  7  shows the  condition that  existed when  the
        tesfcs  were run.  Note  that the liquid  was fed at  a
 i   o  -heat input of 10  MM Btu/hr.  The solids  were  fed at  a
 ih3* fa    lb/batch every 15 minutes (4 times per  hour) .  The
 fl  fitu/iJ had  been  sized to  provide  sufficient  volume, for 20
 JfV VQI r wifch  sufficient excess air  to operate with 10% O2
 B^H u e in  the stack. We  soon found  out   that  the  APC
 to cheB    undersized  due to the  incidents that occured. As
 W8  r*»Were fedf tne draft  in  the  primary  would go positive.
 jjU
-------
resulting  in  the oxygen  level dropping  to 0%  and  a su
-------
   tJlow  varies due  to   uncontrolled flow  into the
                                                      o    e
       Of i8 resulted in a raPid increase in temperature  and
                     Products with the resultant  increase  in
  usino     liquid injection incinerators, solid incine-

not hav/Tv,    °US      With  an auger or  screw feeder
  Is  on* fhV6 sur9,es' The  best control  to minimize
       s on*   h         ,
   8men,  TH-      °Perates at  constant  volume through the
                                by a balance of feed inputs

                   °llred  liquid  feeds  into both Primary and
                  haB W.e11 as a  fast  actin9  Oxygen control
                  heat  input  rates of the secondary feeds.
    Re
     f source Conservation and Recovery Act.  Standards
     Incin     rs  and   Operators of Waste  Facilities:

     ****auly09r*i9i!.C** 264f  RCRA  3°°4f Jan*  25f  1981'


     Ba2a!? flftllfa?Tt0leri'  ."Desi9n and  Operating  Problems of
     
-------
   Air
                                                V»nluft Sciubb«l I  I.D. Fin   SUck
                             FiQ.  i    Rot»iy Kiln lneln«i«IOf
OflOAHIC
                             2. '
                                         456
-------
Ol
                     -J~= ~ ==1
                    *»   |        J
                                     Vl)s   s
                                                                                        EFFLIENI
                                                                                                     Pi jure 4.
                                                                                                    r-AUXHABV BURNER TO
                                                                                                      MATNTAM MMMUM
                                                                                                      COMBUSTION TCMPCftATURC
                                                                                 wgTE
                                                                                 FEED
FBMWT
                                                                                                      VOLATILE CONTENT
                                                                                                      OF WASTE
                           Pig 3
            suawssr
                                                                                                                         §
-------
30
                                                                                                   y Wi+iireit
                                                                                                                •FT;   ®

Botc J
                                                                                                                                                                 -®
-------
                                       Case fi



J

3
CD
i
Ol £
- I
V \l V/' '( » "
\ \ / ' I iJ
i- \1 / ^
i V '/ "i 1Z

4
I l5 3*0 4^ 6*0 °
C
TIME, minutes
. 	 _ 	 _ ^ 	 . 	
	 	 --.....,. 	 ^
	 - - -- . . ^ _
	 	 . . _

~i
/i ii ' :, '' l( (i '( ,. /" /' / , .' 	 — *k*- Capacity
/ •- - , . , Case 'A
' i
• 	 Liquid Waste
10 HH Btu/hr

15 30 45 60
TIME, min.
Solids Loading, 270 Ib. (1,66 HH Btu) every 15 min.

Case A    Volatilizes in 5 min.
          Equiv. Heat Release - 20 MM Btu/Hr.

Case B    Volatilizes in 3 mm.
          Equiv. Heat Release - 33.2 MM Btu/H.

                    Fig. 7

         Effect of Solids  Volatilization
Solids Loading - 90 Ib.  (0.55 HH Btu)  every  5 min.

    Case  A  -  Volatilizes  in  5 minutes
              Equivalent Heat release  : 16.7 HH Btu/hr
    Case  8  -  Volatilizes  in  3 aiinutes
              Equivalent heat  release  : 21.12 HH Btu/hr

           Pig. 8

   EFFECT OF SOLIDS VOLATILIZATION
-------
0)
o
                                                        I

                                                       »!)*>
                                         J.,
                                    IMirr '

                                    If fl I»W
S?J ll/k' 	
o.() m n/dr
                 to.5 (H m/kc




                    _L


                I T|,«*!m mi/iii
  I	*~^_—i
^4 Ti'niw wK/k' j—

 L^	i
                                  ll«. r.rl IIF-

                                  nm tk>r
                                                            •2*

                                                            Kl
                                                     II. (1 m m/kr
                                                           •TO

                                                           Kl
                                                                    TCiS
                                hlU> F..J K.l.t_

                                IXXM Ik/kr
                  ~^7—\
                  . n.» Ih/M U '5 f
j ^
feh* Up V«Ur> 	 • 	 J
M mn
100 C
18 CIB ;
Mrt'e
i
4
0?
Ur
f • 5m«
It"' .
i±J
1 **
"trl
.!_
Ic4
»>55
HCI
Id

TM
I0l»(
Ml
._II
1 b/lir
i<»
5MI
lilTJ
ill
1.21
IOC
r

r it/he
5JO'
I.I
^jr -j1
1
rllri llr
HftlT
tent
1r 1T>t> Ifc
l(t 901
oollxi U*U
7JJ |k/l.t
1 *<•.
»i™. (w*
J| Bin
1 .
I
- i • t
't.t
n
t-
. Drr)
olll)
                                                                                                                                                                               l?*0(.,)
                                                   fiq.  9
                                                                                                                                             Fig.   10
-------
      Dl
        A'  MULTIPOLLUTANT  FIELD STUDY TO  ESTABLISH LEVELS
        CONTAMINANTS IN  AIR,  SOIL,  SEDIMENT,  WATER AND
             PRODUCTS  FROM A  MODEL  MUNICIPAL WASTE
L< *>*>
?'Ha!?kin>'  T-  Hartlage',  R.  Watts',
 -S.  E ess*»  J. F.  Walling*,  D.  Cleverly*
     nvironmental  Protection  Agency
             OH 45268 and  'Research
               NC  27711
 • J
Snc itz9erald,  G. A.  Heil and  H. T. Garabedian
i *t»  °f Environmental Conservation
     °f
     UrX, VT  05676
Ci    ee t>
        "^search Corporation
           OH 45206

         Past,   assessments  have  been  directed  only  at   the public
         8  posed  by  stack   and fugitive emissions of pollutants from
        **ste  combustors  (MWCs)   by   direct   inhalation  exposure.
        the  U.S.   EPA  has  developed  methodologies  to extend  risk
\   dir  t0 a consideration  of indirect exposure pathways in addition to
      eCt   inhalation   pathway.      A  better  undertanding  of  the
            and multiple exposure  pathways  of  emissions   from waste
            needed in  order  to  improve the  overall assessment of the
        humans   and  the environment.    This  study  will   provide  a
         determination of the overall risks associated with NWCs.

»1|Stft(1P*rt  of this  overall approach,   a field  evaluation of a MWC was
V*nt a? Rutiand»  Vermont.  During  a one-year  period,  a four station
O * of y r  ™onitoring  network for  pollutants will be operated by the
V  ion ermont<   The purpose  of the  monitoring is  to ensure that the
SH cientC°ntrols and re9ulatory requirements imposed on the facility are
           prevent adverse  impacts on human health  or the environment.
           8ai"Pies  will be analyzed for specific organic and inorganic
      arf* also for mutagenic components which may  be associated with
      c  itiQn,  grab samples  of agricultural products,  soils, sediments
      a* Vaters surrounding the facility will be collected and analyzed
          toxic contaminants.   This study will provide a framework for
          ollutant,  multimedia field assessments of other MWCs.
   r* Milt
        XlP
                                461
-------
Introduction
                                                                     , ,jt>
     In  the  past,  most  analyses  of human health risks associateo  ^
atmospheric, emissions  from  combustion  sources  have   focused  °ni j,gv'
exposures occurring  by direct  inhalation.  Recent studies,  however*  >t
linked elevated levels of  pollutants in  soils, lake  sediments  and   ^
milk to atmospheric transport and deposition of pollutants from cowboy
sources'-*.  These studies  suggest  that  deposition  of atmosphefi"^
emitted pollutants  could result in additional indirect  routes of e*pgjlid
for humans as well as other organisms.   Recently, the   U. S.  Environ   „(
Protection Agency  (EPA) developed methodologies to extend risk as*6' ^
to a consideration of indirect exposure pathways in addition  to the   ^
inhalation  pathway.     A  better  understanding of the  multipollutadej i'
multiple exposure pathways of emissions from waste combustors is  nee^&-
order to  improve the  overall assessments  of the risks to humans a
environment.                                                           ,
                                                                     if'
     The U.S.  EPA entered into a cooperative agreement   with  the  s  ^
Vermont to conduct a field evaluation of a municipal waste combusto   ^
in Rutland, Vermont.   The purpose of this study is to provide a  *r
for future multipollutant, multimedia field assessments  of MWCs.

The Rutland Resource Recovery Facility
                                                                  Rttll<
     The Rutland  Resource Recovery Facility (RRRF) is located i" * ^ >.
Vermont,  a  city with  a population  of *18,000.   Rutland  is sit g0t *JJ
west-central Vermont in a mountain valley, with the ridges to the ^tQ
west rising over 1000 feet  above  the  valley  floor.    The *flC  Of ^
western Rutland  is located  on rather flat terrain at an elevati"    $
ft. mean sea level (m.s.1.).   Hills rising  to over  1000 ft.  *' ^t^
present to  the immediate north-northwest and south-southwest.  E
over 2000 ft.  m.s.1.  are found 7 km to the east.                       *
                                                                   
-------
    Th-
  M      project  IB a  site-specific and not a  process-specific  study.
   °re» the inorganics  and  organics  may   also  originate   from other
     ln  the area-   The  VAPCD has identified several  pollutants to be
     t* in ambient air during this project:
       Arsenic                       Nickel
       Beryllium                     Benzo(a)pyrene
       Cadmium                       Chlorodibenzodioxins
       Chromium                      Chlorodibenzofurans
       i-ead                          Polychlorinated  biphenyls
       Mercury

    ditionally, a  mutagenicity  bioassay  of  the extractable organic
    n °f the collected ambient air samples will  be conducted.

  «B°!lutantB identified  for quantification  in  soils,  water,  sediments
    lc«ltural products are:
       Arsenic                       Chromium
       Beryllium                     Lead
       Cadmium                       Mercury
       Chlorodibenzodioxins          Nickel
       8 to measure maximum downwind ground  level   ambient air impacts
     - Jcif»erator emissions.   These dispersion models considered source
     lAP  iC8f terrain» meteorological data and receptor  location.   Both
      J*  6 version  of the  Industrial Source Complex Long Term (ISCLT)
      * the LQNGZ  ModelT  were  used  to  predict   average  annual air
      tiona of pollutants in the vicinity of  the MWC.
     *
V  r Se VC*r8  of  ">eteorological  data   (1970-1974)   from  the National
%l*bU rvice Station in Albany, NY, were  used  because the   most  recent
S  d »a *ete°rological  data  for  several   years  in  an area similar to
Vj ttl*  s required-  Modeling was repeated using data  recorded at  a  site
K Ht0n   te Courthouse  in Rutland  and  cloud  cover observations  from
"t'^U  ' VT' during the 1-year period of  1980.  The   wind patterns  were
        *ry  8llnilar to  Albany, with Rutland having  a higher frequency
        H   80uth*est  through  north-northwest.    Using   the Rutland-
        data, results similar to the analysis with the Albany data  were
 •*iv '"mi   """"* *ere run using polar grid receptors   as  well   as discrete
 S '*tin-fecePtors.   Maximum annual average ground  level  concentrations
 l<0 '^Ctd     fro» the source for 16 wind directions  beginning with south
         »v*ry 22.5°  along the polar azimuth at distances  of 0.2,  0.5,
          '  10,  20, 30, 40, and 50 km from the MWC  (for a total  of 160
           The discrete  receptors were sited at 59 locations to better
        p°int  of  maximum  concentration  and  were  also   placed near
        **9nents  of  the  population,  e.g.,  schools and  hospitals.
         °d»ling showed the areas for maximum impact lie  within  a 1 km
                                 463
-------
radius from the MWC stack.
Air Monitoring Sites

     Based on  the results   of  the air dispersion modeling, a four-station
ambient air network was established for monitoring the selected metals and
organics.   The stations  are located  on municipal property with three of
the stations close to  modeled  sites  of highest  estimated annual average
concentration of  pollutant  emissions  or close  to areas of topographical
importance, and the other station  is the existing Vermont State  and Local
Air Monitoring Station (SLAM) shelter.   The sites are (Figure 1):

     1. State  APCD SLAM  Shelter:  1.3 km east in an urban area of Rutland
        City at the State District Court House.
     2. Old Havenworth School on Watkins  Avenue,   Rutland  City:  0.25 kit
        north near  the modeled site of estimated maximum impact from the
        facility and located in a  residential neighborhood.
     3. Rutland Town Municipal  Building: 0.7 km west in  the narrow valley
        formed by Otter Creek.  The site is 200 ft to the south of Route 4
        behind the Rutland Town Municipal Office Building.
     4. Next to River Street Pumping Station,  River Street,  Rutland City:
        0.4  km  south-southwest   in  a  topographical depression near the
        junction of East Creek  and Otter Creek.

     Two ambient air  monitoring   stations  have  been  designated  as co-
located sites for quality assurance purposes8.

     At each  monitoring network,   several types of sampling equipment are
being used.    Each monitoring  site  has  two  General  Metal  Works PS-1
samplers, one  standard mass flow TSP  (Total Suspended Particulate) high
volume  (Hi-Vol)  sampler  and  one  Wedding  PM-10  critical  flow Hi-Vol
sampler.    The  Hi-Vol  samplers,   TSP  and  Wedding Model PM-10,  and the
Pesticide Sampler Model PS-1 are  being  employed to  monitor the selected
metals and  organics in  the ambient air surrounding the MWC.  The metals,
arsenic, beryllium, cadmium, chromium,  lead,  nickel and silver,   and B(a)P
will be collected by Hi-Vol  PM-10  samplers.   Separate PS-1 samplers with a
glass  cartridge  filter  and   polyether-type  polyurethane   foam  (PUT)
adsorbent  are  being  used  to  collect chlorodibenzodioxins/chlorodibenzo-
furans  (CDDs/CDFs), PCBs and a  sample  for  bioassay  analysis.     The TSP
sampler with  mass flow  controllers and  high quality glass fiber filters
are used to collect particulate for  determination of  mutagenic activity.
Ambient air  samples for  mercury  are collected,  using a low volume vacuun
sampler with a mass flow controller,  only at the SLAM site due to the need
for a controlled environment for the collector.

     The sampling frequency  for the PS-1 PUF samplers, the TSP high-volunte
samplers, and the PM-10  high-volume samplers  will be  for 24  hours once
every   12   days  for   approximately  1  year.     This  will  produce  26
samples/collector (one collector at each site and a  co-located collector)
or 130  air samples for metal,  organic chemicals for mutagenicity testing,
B(a)P and PCB analyses.

     The number  of  ambient air   sampling  days  for  CDDs/CDFs  will be
restricted to  21, producing 105  samples from the 5 monitors.  The tetra*
through octa-CDD and CDF congeners will be the isomers to be determined in
                                    464
-------
tk» a»bi
lot ^ •>lent   air.   The five other sample days on which the PS-1 sampler is
   ect   U8ed f°r CDDs/CDFs data collection, ambient air samples  will be
     ed and analyzed for mutagenic activity.  Mercury sampling (24-hour)
  dftyly8ia wiil  be conducted by the VAPCD at the Rutland SLAB site every
   y
>r* COM  *°r  *^e   duration of the study producing 26 samples.  All samples
    u*cted on  the same day.

       Qrol°9ical  information,   i.e.,   wind  speed,   wind  direction and
          are continuously monitored and recorded at two sites, the SLAM
      at»d  site  f2   at   the  old  Havenworth  School  on  Watkins Avenue.
 -  » ni°8 Eiectronic   Weather stations are used to continuously measure
Xf,J*teorQlogical   variable.     Additionally,   the  SLAM  site collects
  '**ti   int*nsity,   relative  humidity,   atmospheric pressure, and solar


     ^ RRRF began burning municipal waste in November,   1987.   Figure 2
           relationship between  the tons  of waste  burned and the air
           of  particulates at the four air monitoring stations.

                        Analytical Procedures
   A*E
          and chromium  in ambient air are analyzed  for total  metal by
 I1 l be 8ctivation   (HAA)».    Beryllium,  cadmium,   lead,  nickel and silver
 »i  A£S)!7iysed by   an  Inductively  Coupled Plasma  Emission Spectrometer
 f»  *iect  *  PCBs in ambient air will be quantified by gas chromatography
 Olu*io     capture  detection  (GC-ECD)11.     High  resolution GC-high
 J°u9h  n  HS  P   will  be   analyzed  by  thin-layer
 V/**t»j! %Phy  and    fluorescence  spectrophotometry1".      A  pyrolyzer-
  cHfy v Syatero has  been   developed by  VAPCD to  quantify both elemental
       P°r and total  mercury.
               Of  Ambient  Ai
      fr°Uiate  matter   on   TSP  high-volume glass fiber filters and PUF
        h?  periods not  used  for CDDs/CDFs ambient  air analysis  will be
       ch    88y  the   collected  materials for mutagenicity. Individual
       i einlcai8  will   not   be  separated  and/or  identified,   but the
         0l>Banlc  fraction  from  the filters,   most likely a mixture of
 \ **  P*  !Und8' wil1  be  test*d  for mutagenicity.    Mutagenicity testing
 ^1   Qr th°riBed by  either  the   standard Ames  test plate incorporation
    9fOup    Kado assay", a modification of the Ames test.   Samples are
    i(i|f f    for bioassay analysis in order to minimize test variations
          r Accurate comparison  of  sample sets.
   I
 r  t"*  Of" Wodelin9 °f  emissions from the Rutland RRF  indicated that
 Slh>'c!UjCtUty*Xpected  maximal  deposition  will be within a 1 km radius of
      r>m      Locations   for   collecting  water,   sediment,   soil  and
      P1*s Product  samples are   generally within  the high-impact area.
      CUy H  *ater  and   sediment   are  taken  at   five  locations: the
         t8  8ervoir, Rocky Pond and at three points in the Otter Creek.
          ' Potatoes and  forage grass  hay samples  are being collected
                                 465
-------
                                                                     c
from farms  in the  area surrounding the  RRRF.   A  systematic grid saw3 ^
technique is being used to collect soil samples  at  the  five sites-  ,t
metals in  water, soil, sediment, milk and  agricultural products are  '^
measured by atomic absorption"'17.  The  analysis  of PCBs,  and PCDDB'
will be conducted using HRGC/HRMS'•-•«.
Conclusion

     This  study  is  the  first  to provide  scientific evidence
ground level concentrations of air  emissions  as  well  as  depos
other  environmental  media.    Protocols   are   being  developed to f
guidance for future field evaluations of municipal  waste combustor**
References
                                                                     tr
     1.   MEA, Inc., East Helena  Source  Apportionment   Study,  "P8f
         source  apportionment  using  the   chemical mass balance   ^
         model*, Volume 1,  Prepared  for   the   Department  of  H*
         Environmental Sciences, State of Montana,  (1982).

     2.   D.  L.   Swackhamer,  "Estimation   of   the  atmospheric  f"
         atmospheric contributions and loses of polychlorinated
         for  Lake  Michigan  on  the  basis of sediment  records °
         lakes", Environ.  Sci. Technol. , 20: 879 (1986).
      3.   J. M. Czuczwa,  R.  Hites,   "Environmental   fate  of
          generated  polychlorinated   dioxins   and furans",
          Technol. 18: 444  (1984).

      4.   C. Rappe, M. Hygren, G. Lindstrom,   H.  R.   Buser,  0.    ,
          Wuthrich,  "Polychlorinated  dibenzofurans  and dibenzO'P
          and other chlorinated  contaminants   in  cow  milk  *f   i\'<
          locations  in  Switzerland",     Environ.  Sci.   TectjB$&'
          (1987).
                                                                tie*0*
      5.   Air Pollution Control  Division,   Agency of  Natural  tftt
          State of  Vermont,  "A  review of the potential health  ^ rfi
          dioxin emissions, acid gas   emissions,   and  disposal   ^
          contaminated  ash   from  the Vicon   resource    cove
          proposed for Rutland, Vermont,  (1985).
                                                                0
      6.   U.S. EPA,  "Industrial  source   complex   (ISC)   disp*£  ^
          user's guide",  (2nd edition),   Research  Triangle Par '
          450/4-86-005a (June, 1986).                               „/
          U.S. EPA, "User's instructions  for  the SHORTZ and
          programs", Philadelphia, PA, PB83-146100 (March,  I98
      8.   Air  Pollution  Control  Division,   Agency of
          State  of  Vermont,   "Rutland   resource  recovery   
-------
   trace  elements   in   suspended  partlculate  natter collected on
   glass-fiber   filters*,   Office  of   Research  and  Development,
   Environmental Monitoring   Systems Laboratory,  Research Triangle
   Park, NC, SOP-EMD-017 (1984).

 Ot U. S.  EPA,   "Standard  operating   procedure  for   the  ICP-DES
   determination of   trace elements in suspended particulate matter
   collected  on glass-fiber  filters",   Office  of  Research  and
   Development,   Environmental   Monitoring   Systems  Laboratory,
   Research Triangle Park,  NC,  SOP-EMD-002 <1983>.

ll< U.S. EPA, "Compendium of methods for the determination  of toxic
   organic  compounds  in  ambient  air".   Office  of  Research and
   Development,   Environmental   Monitoring   Systems  Laboratory,
   Research Triangle Park,  NC,  EPA-600/4-84-041 (1984).

 2> R-L.  Harless and  D.   McDaniel,   "Method  for determination of
   Polychlorinated dibenzo-p-dioxins and dibenzofurans in stack gas
   emissions  and   ambient  air",    presented  at:  1988  EPA/APCA
   Symposium on  Measurement of  Toxic  and  Related  Air Pollutants
   (I*ay, 1988).

  * U.S.   EPA,    "Standard   operating   procedure  for  ultrasonic
   extraction and analysis of  residual benzoEaJpyrene  from hi-vol
   Alters via   thin -layer chromatography",   Office of Research and
   Development,  Environmental Monitoring   Systems      Laboratory,
   Research  Triangle  Park,   NC,   EMSL/RTP-SOP-MDAD-015 (December,
   1986).

   D-N. Maron and B. N.   Ames,  'Revised  methods for  the Salmonella
   "•"tagenicity  test",  Mutat.  Res..  113:173 (1983).

   "•Y. Kado,  D. Langley and E.  Eisenstadt,  "A simple modification
   °f  the Salmonella liquid-incubation assay.  Increased sensitivity
   *0r  detecting mutagens  in  human  urine",  Mutat.  Res..  121:25
    [J'S. EPA,  "Methods  for  chemical  analysis  of water  and wastes",
    ffice of  Research  and  Development,  Environmental Monitoring and
    buPport Laboratory, Cincinnati,  OH,  EPA-600/4-79-020 (1979).
I?
    L'K* EPA'  "TeBt methodB for  evaluating  solid waste.   Volume IA:
    rjaboratory  manual  physical/chemical  methods",  Office of Solid
    *8t* and  Emergency Response,  Washington,  DC,  SW-846 (1986).
la-  U q
    '=  EPA,   "Methods for  organic chemical   analysis of municipal
    g   industrial  watewater",  Office  of Research and Development,
    Epiironi"entaJL Monitoring  and Support Laboratory,  Cincinnati,  OH,
     A-600/4-82-057  (1982).
    U Q
    «Uh   EPA'   "Protoco1   for  th* analysis of 2, 3, 7,8-teratachloro-
    r«.*?Z°~p"dioxin  fay   high-resolution  gas  chromatography/high-
                mass   spectrometry",    Office   of   Research  and
         Pment, Environmental   Monitoring  Systems  Laboratory,  Las
          NV, EPA-600/4-66-004 (1986).
    U.s
    »nd*       "Analysis for polychlorinated dibenzo-p-dioxins (PCDD)
        d*benzofurans   (PCDF)   in  human  adipose   tissue:  method
    EVaion    study",    Office   of  Toxic  Substances,   Exposure
        u«U0nB  Division,  Washington,  DC,  EPA-560/5-86-0 (1986).
                             467
-------
00
-------
   SAMPLE DATE
Figure  2.
                                                                   ROUTE A
                                                                   UG/CUBIC M.
                                                                   SLAMS
                                                                   UG/CUBIC M.
                                                                   RIVER ST.
                                                                   UG/CUBIC M.

                                                                   WATKINS ST.
                                                                   UG/CUBIC M.

                                                                   TONS BURNED
             20    40    60   80   100   120  140   160   180   200   220
Relationship  of  Tons of Waste  Burned and Particulate
    Concentration  at the Four Monitoring Stations.
-------
                          ABSTRACT

                       Seong T.  Hwang
                       Exposure  Assessment Group
                       U.  S. EPA
                       Washington , D. C.
             Assessment  of  2,3,7,8-TCDD Emissions from
                  Waste  Disposal  Sites

     The  public  is increasingly  concerned about the public  h* j»|
 consequence  associated  with  sites  spilled with wastes cone3*
 2,3,7,8-tetrachlorodibenzo-p-dioxin(2,3,7,8- TCDD). AlthougJ^
 2,3,7,8-TCDD  exhibits extremely  low vapor pressure, inhalat -i'
 exposure  resulting  from emissions  from the contaminated **'  f>'
 one  of the pathways  which  will  affect  the human health t^a j i(,
 emissions could  be  in vapor  or  particulate form. In addit* j(Ji
 the  health risk  consequence  associated with spill sites* *'  ,
 TCDD containing  wastes  being generated and disposed of a*  wa«'
 result of production activities  or site cleanup operation*
 also lead to  airborne emissions  and consequently inhalat*0
 exposure. In  this paper, the significance of 2,3,7,8-fCD0  j If
 emissions from  the  spill and disposal  sites will be ass**
comparing the exposure  to  the emissions with the expos^r*  f
associated with  other multimedia pathways. The potential  .
human health  risk will  be  examined for various scenario*  jy*
environmental conditions and disposal  design at differe°C
of 2,3,7,8-TCDD  contamination in the wastes.
                              470
-------
   °ioxins  refer  to  a  series  of  related  chlorinated  compounds


  ying  from tetrachloro  to  octachloro-compounds with a

8t^ct
    •-urai backbone of  dibenzo-p-dioxin.  The  compound that has


    a subject  of  many  studies among  these  compounds  is  2,3,7,8-


   achlorodibenzo-p-dioxin(2,3,7,8-TCDD) because  of  its extreme


    and chronic  toxicity observed in  animal studies. 2,3,7,8-


       become a  household  word  by often referring it to as


     dioxin implying  that  this  particular constituent  would be


       concern among all the  dioxins.  This paper  will


         _ use the  term dioxin  to denote  2,3,7,8-TCDD  unless

 ^tWi«,« o
    *8e described.

   DioyJ^
     *j.n is formed  during  the manufacture of the


    °raphenol precursor. A good example of  this  process is the

 ^Uf »

    °ture  of  a herbicide,  Agent Orange, which is formulated to

 ^fcfti
>      2'4»5-trichlorophenoxyacetic  acid(2,4,5-T) as the major

  iv» in
       "9redient. Dioxin is a minor  contaminant formed  as a


     ct during the  synthetic process. This  illustrates an
   IV*4«

        Principle of  evaluating commercial  products: Exposure


       n€ of a product must not  only be concerned with  the major
   'Oh

  %       but alfl° with  the  minor contaminants that may be


      *8 * result of their  formation in the manufacturing


      °r as a result of transformation reaction occurring in

      rcm
       °n«ent. This consideration is also true in evaluating
                            471
-------
 the  stack  emissions  from  certain combustion processes where
 dioxin  is  reported to be  present.
     Occurrence  of environmental pollution by dioxin began
 improper disposal of contaminated wastes (Times Beach, Mis
 Eglin Air  Force Base) ,  spraying of contaminated pesticides,
 accidental releases  of  the contaminants during manufacturing
 procedures (New  Wark, New  Jersey; Seveso, Italy),  and ubicp1^
                                                           6.6**
 presence of this contaminant  in some of the combustion proce
                                                            i"
 The  importance  of dioxin  as an environmental pollutant lifiS
                                                            t$
 its  extreme acute effects and chronic carcinogenic potency
                                                             01
 in animal  expseriments ; capability to detect trace quantit*
 dioxin  in  the environment; magnitude of bioavailability wh*
a function of presumably the  soil matrix  effects;  and ina&
                                                         flU** *
to elucidate the toxicological  significance  of  human expo8

the contaminant.    This paper  is mainly  concerned with
                                                            < 0
evaluating exposure potentials  of dioxin  emissions from SP

waste disposal  sites. Exposures associated with the     8
from the combustion processes  are  not considered in thlfl
    If soil contamination  is prevalent,  the emissions can    ^
the form of vapors and particulates  as a result of
                                                         flt*
and wind erosion, respectively.  Although several experi*
measurements have demonstrated  that  dioxin exhibits an
low vapor pressure, some  studies  have shown that the &°6
                                                      Ai6t
important loss mechanisms occurring  in soil or waste <**
                                                       ion
sites are assessed to be  volatilization,  and soil eros1
                                                       A
winds or runoff (Freeman & Schroy,  1986;  Young, 1983) •
                              472
-------
  Qy has  shown  that because of a  large soil  water  partition
   "icient  the  amount  of dioxin leached by water out  of  the  soil
  e*tremely small(U.S.  EPA,  1985a).  Photolysis  may occur on  the
 Peca»ost  surface of the soil layer upon which sunlight is
   ^nt,  although  at a very slow  rate.  But the  bulk of the soil
ih
   ft*ch the majority of dioxin is present will  be  unaffected by
   °lysis.  significant degradation of  dioxin by microorganisms
   oil is considered highly unlikely (Matsumura &  Benezet, 1973;
     1983).
     Ce dioxin  becomes airborne in the form  of  vapors or dust,
       r  human  exposure occurs through inhalation, a  direct
       Soil erosion by runoff will  form sediment  in  water
     *nd impact aquatic organisms or  water  quality.  Human
     re through ingestion of contaminated aquatic  organisms  is
*n
          pathway. Other  indirect pathways associated with
 nt
Sl        — sites  include  the  contamination of the food chain
 "oh „
    ftfi toh
^     'Beat or dairy products.  This paper will compare the
   ^•tUrto
Ht         exposures  associated with vapor and dust inhalation
    Othftv.
fc      r exposures associated with the indirect pathways, if
 *irt-
   MHY%4» J
1 ingestion  by children and dermal contact with the

*H    c°ttParison  of exposures can be presented in terms of
       Banking for the pathways occurring on-site under some
    *1
      Patterns of human activity,  but such a comparison is not
                            473
-------
 straightforward  for  exposures occurring to the population
 residing  some distance away  from the site. The former expo3
 occur  onsite, while  the  latter occur offsite. This paper
 also describe the methodologies used in such exposure
 assessments  and  discuss  the  factors that can influence such
 assessments.
VAPOR  INHALATION
    Evaluation of exposure associated with vapor inhalation
                                                         of
based  on  first estimating release rates and then the use «"•
type of fate and transport models. In the assessment    o*e
                                                          .
here, onsite and off site exposures are separately estimate •
                                                         te
estimation procedures and the  fate and transport models a*
from published  information. Release rates are estimated
                                                          pX
considering thermodynamic equilibrium conditions between  f
                                                         j F*
and the transient process of concentration changes (Hwang
1986) . Evaluation of the transport of the contaminant
point of release to an onsite  or offsite point of
based on the concept of wind dilution of vapors partition
                                                         Ol»
between soil pore air and soil, which emanates from the
the Gaussian dispersion model (Wark & Warner,  1986), resp
    The vapor inhalation pathway potentially  active  at
will largely be a function of the physical and  chemical
properties of dioxin, local environmental conditions^0"   fl
                                                        c*
winds), and current and projected local  land  use.
                              474
-------
   Ptor and the dioxin concentration at the site is not a
   Dining factor in relative ranking of onsite exposures among
   Pathways because the onsite vapor concentration is not
      to be a function of distance in contrast to  the offsite
      °n, and because the dioxin concentration in soil linearly
      ambient air concentration and cancels out in the relative
      process.
    Pecific procedures and models used in the release rate
       and the fate and transport analysis are not enumerated
in *M«
      Paper. Chemical and physical properties needed in these
       can be found from the references cited in this paper.
Sr
    Of
      these properties used in this assessment for relative
     9 of importance between various pathways is listed  in Table
  Th
     8e Properties are chemical and media specific. Care must be
     ed in applying some of these properties to situations
        different types of soil medium.
            Table. 1 Chemical and Physical Properties Used  in
                       Dioxin Volatilization Rate Analysis
           Partition Coefficient: 4680 kg soil/L water
            Constant: 1.6xlO"5
          Coefficient: 0.05 cm2/s
^w    lation of wind borne contaminated particles  results  in
  °^t« 4.
l*v        dioxin because the particulates may contain the  same
     Of H4
              as in soil at the site. The extent that
                            475
-------
particulates will become wind-borne  is a complex  function of
particle size, relative humidity, particle chemistry, topol°#'
and wind velocity. Particle size also determines  the  lifeti**
the particle in the body(inhaled versus swallowed). A recent
                                                            jii«'
publication designed  for rapid assessment of the  wind-blown
describes a procedure for  release rate analysis(U.S.  EPA/
    Once the release  rate  analysis of dust blown  by winds *•*
complete, the fate and transport analysis follows the same
procedure used for estimating exposure to volatilized dioX*n<
                                                          6a6*
Here again, the estimation of absolute exposure  is not  nee*
for comparing the relative importance of the exposures.  AS
                                                         sur«
mentioned in the previous  section,  the  importance of
associated with the dust inhalation pathway in relation to
exposures from other pathways  can be compared for onsite
                                                         ntX
exposures without  an explicit  assumption about the concenv
level at the site  and without  reference to distance from
                                                           -g
to the exposed receptor.  But  it  can be a function of
from the source to the  receptor  for off site exposures
comparison of exposures for offsite pathways can not be *"
until certain assumptions are made about the receptor
INGESTION OF MEAT AND  DAIRY  PRODUCTS                        J
                                                       rtil »' r
    Grazing livestock  ingest a  substantial amount of s0i    ^
of their dietary intakes. Also  livestock grazing in past
                                                        to-, I*4'
contaminated site will ingest the  contaminant taken up °*   ^
plants they eat or deposited on the leaves following
                              476
-------
Of
     dust, soil contamination can occur on offsite location
      land when the wind-raised dust at the contaminated site
     :a on the offsite land, or when soil erosion due to
   *Pitation runoff accumulates the contaminants at the offsite
       Ln being ingested by livestock may bioaccumulate in the
  8Uea  of these animals or may be excreted  in milk in lactating
  *atock, or may eventually enter the human body when meat
   u°ts are consumed as food. The magnitude of human exposure to
     will be dependent upon the amount  of  soil and plant
     1  »nd the dioxin concentration  in soil. As mentioned
     laly, the dioxin concentration  in soil is kept at the  same
     Uaed for comparing exposures resulting from all the
       Pathways.  There is an emerging body of experimental data
         the extent of bioaccumulation  in  the bodies  of  animals
 *^U
        contaminants through soil or plants. An equilibrium
     °nship(Fries.  1985)  can be assumed between the dioxin
  nceni-
     Cations  in animal  meat or milk,  and soil, or transfer
   **t
     °ients can be  used  to estimate the concentrations in meat
       if  the  intake rate of dioxin through the soil and plant
          be estimated(U.S.  NRCf 1982)
          ION
      in9estion is common in children.  Exposure may occur
   *t
      *B a result of intentional eating of soil, as the result
            state (pica), or through inadvertent ingestion of
                             477
-------
                                                            jjd
soil as a result of mouthing soiled objects.  While there i&
consensus in the literature regarding the precise age range
during which such behavior occurs, most investigators defin6
range as between 1 and 5 years of age.
  When soil ingestion occurs, contaminants present in the 6°
will enter the gastrointestinal tract, and may be absorbed *
the body.  The extent of absorption is dependent upon many
factors, including the biological uptake mechanism leading
absorption, the metabolic state of the body, the physicoche"
properties of the chemicals present, and the degree of ads'
on soil. Bioavailability of dioxin in soil is less than in
unadsorbed state. Absorption of dioxin on soil through the
gastrointestinal tract is assumed to be 30%.
DERMAL CONTACT
  Individuals may come into direct contact with soil in a
of situations, such as when children play in a yard or
adults work in a garden planted in contaminated soil.  De**  ,
contact will not result in systemic toxicity unless conta*
are absorbed into the body.  To determine dermal absorpt*0
                                                           $
needs to know the concentration of contaminant in the so*1'
                                                          jP
amount of contaminated soil in contact with the skin, the
chemical absorption through the skin, and the duration an
frequency of exposure.                                      ,
                                                        A &
  The total exposure for direct contact with contaminate0   ,
also a linear function of the dioxin concentration in
                              478
-------
  *tive ranking between  onsite  exposures  from various pathways


   a*>solute concentration  needs not be known. Dermal absorption


      in soil  is  assumed to  occur  at  a rate  of  0.5% upon  contact


       skin in this  assessment.





  ER EXPOSURE  PATHWAYS


        are several  other pathways of importance  in comparing


      s due to dioxin at  a  contaminated  site.  Dioxin exposure


     occur through  ingestion of vegetable grown on the  soil,

in9est<
    *"*on of fish  affected  by the contaminated medium, and
ijjg

     i°n of surface water  which received  sediment  eroded  away


 081 ^e site.


     though there have been  some studies  reporting distribution


     *in between  soil and  vegetables  or plants, existing

 *°tia
     ation on  plant uptake is contradictory  and insufficient


     ftich to make a generalization regarding the degree of


      *ot this reason,  the  exposure  due  to  vegetable intake is


     cluded as part of this  comparative assessment.


     °*in adsorbed  on soil or present at  the waste site may be
 *&h

8     tted to  a nearby surface  water  body through  the process of


j     °sion and the runoff of excess  precipitation.  The process


4v   ^ c?»plex and  highly  variable, hence estimating a  long-term


to   9* rate of ^noff is difficult. Soil  erosion will  result in
        of sediment at  the  bottom of the water body.  The

  ***!
      ated sediment along with the runoff will impact the

  *
      water quality and affect aquatic organisms including fish
                            479
-------
 living in the water.  The extent of bioaccumulation of sediment
 dioxin in fish is poorly documented in the literature. When
 contaminated sediment is involved, it appears according to some
 literature information  that bioaccumulation is more important
 than bioconcentration. In addition, exposures due to these
 pathways are difficult to compare on a generic basis without
 specifying some pathway specific conditions for evaluation.
 Although some or all  of these exposure pathways may be extremely
 important,  they are not included in the relative ranking scheme
 on a generic basis.
 EXPOSURE DURATION
     Exposure duration represents a length of exposure to the
 contaminant at the site.  Exposure duration will likely be a
 function of the specific pathway under consideration and the mass
 of  the  contaminant present at the source.  For comparative
 assessment,  all exposures are calculated based on the same length
 period,  possibly lifetime exposures.  Adjustments to other
 exposure period are not necessary in comparing the magnitude of
 exposures as  long as  the  exposure periods considered are kept the
 same  for all  the pathways considered.  The mass at the source is
 also  kept at  the same level  for all the pathways evaluated. Since
 soil  ingestion occurs only during the childhood, lifetime
 exposure due  to soil  ingestion represents an exposure occurring
during childhood averaged over an individual's lifetime.
Exposure duration associated with  fish consumption can be
considered differently than  lifetime exposure.  Literature
                              480
-------
°SURE
   a^y reports  general  consumption for  the  population  as  a
  •Le*  Any  effects  resulting from such  generalized  consumption
    be diluted  by the availability of fish  from a local  source
        by the site.

        ASSESSMENT PARAMETERS
       ^ are many  exposure assessment  parameters needed for
        relative  importance of exposures associated with the
&athw»,
    Qys. Examples  include breathing rate, consumption rate of
^jia
    ytown  vegetables,  consumption of meat from domestic
     °ck.  etc. Space in  this paper does not allow  an exclusive
       of  default  exposure parameter values used in this
           assessment. Generally accepted default  values are
    ln this assessment.  These can be found in some of the EPA
    cations dealing with exposure assessment(U.S.  EPA,  1984;
    EPA, 1986).  Additional details on  calculational procedures
      ^•e»  input  parameters needed, and exposure scenarios used
    e  found in a recent  EPA document on dioxin ready for
    Cation for external  review(U.S. EPA, 1988).

   1/Ts
     * ^sult of analysis as indicated above show that the
      •*P08ure to  dioxin via inhalation of vapors and wind-borne
otv  °UlateB «e  less important than exposures associated with
  *t n
^   Pathways, it  is not surprising to obtain these results
       dioxin tends to bioaccumulate in upper trophic levels
                           481
-------
 in the  food chain due to highly lipophilic properties of
 Relative ranking shows that for the onsite exposure based on
 same level of dioxin concentration at the site, the orders °*
 importance are  l)beef meat ingestion, 2) dermal contact, 3)
 ingestion of dairy products, 4) soil ingestion, 5) vapor
 inhalation, and 6) dust inhalation.
    In  the case of offsite exposures, the inhalation exposu?
 decreases proportionately as the receptor distance from th*
 increases. The  relative ranking of the  exposures  from the
 pathways remains the same. Since offsite contamination is a
                                                             Iv
                                                          fjt\r
 result  of transport of dioxin  from the  original site, the °*
 exposures for the all the pathways are  slightly less than &
 onsite  exposures.
                                                            JF
    Onsite exposures to dioxin through  inhalation of
 airborne dust are estimated to be about 1x10"^ and 5x10"
 ng/kg.day, respectively, when  the level of dioxin contain
 at the  site is  1 ppb. These exposure levels  represent the
 amount  of dioxin being contacted externally  with  the coflp0
                                                        . i*e>
 of the  body per unit weight of body averaged over a lifev*   ^
                                                       i    «n
 Chronic risk in terms of the oncogenic  effects can be obt*
multiplying the potency slope  by the exposures provided *p

DISCUSSION
    Although the vapor pressure of dioxin is extremely
exposure due to vapor inhalation is greater  than  that
wind-blown dust under an average wind condition observed
                              482
-------
»11

    **    speeds  will  result  in more  particle  entraimnent  from
   site.  This  may change  the  relative  ranking  between vapor  and

    inhalation exposures.  In  any case,  exposure  assessment must

   l<*er other  pathways  such as  beef ingestion, dairy products

   ation,  soil ingestion,  and soil  dermal  contact.  These

    ures  may pose greater exposure  than the inhalation  exposures

   "Qse pathways  are relevant.  Although the present comparative
          did  not  include sediment contamination and the

    ting bioaccumulation in fish,  and vegetable uptakes of

       these pathways  could also be very important exposure

       ,  surpassing the inhalation pathways in its importance.
V
h'3/?*?'  R>  A-'*  Schroy,  J.  M.(1986)  Modeling the transport of
 ^itA TCDD and other low volatility chemicals in soils.
    °n-  Prog.  5(1):28-33.
 *UB
rWtaiv?*  p'  (1985)  Bioavailability of soil-borne polybrominated
  '565*5   in9ested by farm animals.  J. Toxicol. Environ. Health


       •  T.;  Falco,  J.  (1986)  Estimation of multimedia exposures
      to hazardous waste facilities. In: Cohen, Y., ed.
         in an multimedia environment.  New York, NY: Plenum


^CciT"' F>;  Benezet,  H.  J.  (1973) Studies on the
  ^achi   atiuon and aicrobial  degradation of 2,3,7,8-
^    ni°todibenzo-p-dioxin.  Environ. Health Perspect. 5:253-258.

  lt0?JA~ (1984)  Risk analysis of TCDD contaminated soil. U. s.
  ,  Cental  Protection Agency,  Washington, DC. EPA-600/8-84-


                                             annual research
                                             in contaminated
        EPA-600/9-85-013.
                             483
-------
U. S. EPA.  (1985b) Rapid assessment of exposure to
emission from surface contamination sites. Office of Health
Environmental Assessment, Washington, DC. EPA-600/8-85-002.
PB85-192219/AS.

U. S. EPA.  (1988, March) Estimating Exposures to 2, 3,7, 8-TCDD
Office of Health and Environmental Assessment, Exposure
Assessment Group, External Review Draft, Washington, D.C.
600/6-88/005A.

Wark, K.; Warner, C. (1981) Air pollution—its origin and
control. New York, NY:Harper and Row Publishing Co.

Young, A. L. (1983) Long term studies on the persistence
movement of TCDD in an national ecosystem. In: Tucker, A.
Human and environmental risks of chlorinated dioxins and
compounds. New York, NY: Plenum Publishing Corp.
                              484
-------
                                 Prtv«tlint wind*

.
        Coupling between  source location, envlronaental  transport
      «nd huaan exposure pathways.
                            485
-------
THE  IMPORTANCE OF  PROPER SITE  CHARACTERIZATION
OF THE CONTAMINANT PATHWAY
C. E. Schmidt
Radian Corporation
10395 Old Placerville Road
Sacramento, CA  95827
     The Comprehensive, Environmental  Response,  Compensation and L
Act of 1980  (CERCLA) established  a  national  program for responding ^ •
release of hazardous substances in  the environment.   U.S.  EPA est ..
procedures for implementing  this  program,  commonly  referred  to as t
remedial investigation/feasibility  study  process (RI/FS).  This Pr°
outlined in  the National Contingency Plan (NCP),  50 Federal  Regi^te ^eV
5862, February 12, 1985, 40  Code  of Federal  Regulations (CFR)  300.    ^
framework for remedial  response can be described as a five-step Pr.  jji'
includes:  site discovery, preliminary assessment,  establishing P11  j
for the remedial action, remedial investigation/feasibility  study.   ^
remedial action design  and implementation.   Perhaps the singularly  gjpcx
significant  component in this process  is  the remedial investigati°  ^\*
the investigation collects all data needed to identify,  select,  a°
ate remedial action alternatives.

     The focus of this  paper is to  illustrate the need to  conduct
site investigations that accomplish the objectives  defined in  the
on Feasibility Studies  Under CERCLA, June 1985,  namely,  to assess
nature and extent of the contamination and provide  the informati°   j.
for remedial design work.  This paper  will describe technical  appr ^
that facilitate the design of site  investigations particularly f° ru)^"!
contamination pathway,  that  meet  these objectives and provide  int $$!>eA
on survey techniques that can be used  in  a inultiphased approach    p tf
the type, level, and extent  of contamination.  Case study  informs  ^ (^
be presented that demonstrate the detrimental effects that result  Jfl.
                                                                 * ^u  mAf
incomplete site characterization  and illustrate  the importance o*   pi
ing a proper remedial investigation.   The  discussion of  these  det
effects include the obvious  impacts to the site  cleanup schedule ^fi•
budget, but  it also includes the negative  impacts to the on- ana
health and safety program, impact to the  community  via the air P
impact to the waste transportation  and treatment/disposal, and
tial for increasing short- and long-term  liabilities.
                                    486
-------
            THE IMPORTANCE  OF  PROPER SITE CHARACTERIZATION
                      OF THE CONTAMINANT PATHWAY
                              C.E.  Schmidt
                               C.C. Mecham
                              M.T.  Galloway
                          Radian Corporation
                     Sacramento, California  95827
    nil
°n t-L e success or failure of hazardous waste  site cleanup depends largely
            design and implementation  of  the  site  characterization
          namely the site historical survey,  preliminary assessment,  site
        <  and the remedial investigation.   Incomplete  site  inspection and
  •«J,y    lnvestigation , including  the air  contaminant pathway can nega-
tive 3   ct tne feasibility study, selection  of  feasible remedial alter-
ift..   •  and site cleanup.  A list  of the more  obvious potential  negative
        ° the remediation program  is provided  in  Table  I.
    Th'
        n"ill describe types of activities  necessary for proper site
               and one simplified and  general  approach to designing and
         site characterizations that can  assist  in accomplishing the
        te investigation objectives.  Attention has been given to the air
     •o^° that air pathway analyses (APA)  can  be  included in the remedial
         °n.   This approach involves designing a multiphased sampling and
         program that employs a range  of  techniques, from survey methods
          and detailed analyses.  This approach, combined with  guidance
        °n sampling and analytical protocol  and completeness (defined by
        al evaluation of the data), can be used to  perform successful site
        tions that support hazardous waste site mitigation.
lift   C3oa
  89tiv .    study that demonstrates how  an  incomplete site  investigation
       y affected the site cleanup will be  presented and discussed.

                 Designing the Remedial Investigation
(jv  The
41 tacte  .Pose of the remedial investigation  is  to provide  an adequate
te^^af1Z;ation of t*ie site so fckat the most cost-effective  remedial
t *^ati0 6 catl ke developed and implemented.  Design of  the  site charac-
Dtv^6ctiot, technical approach begins during the  scoping  phase,  with the
         stid evaluation of available site information.   Hopefully,  the
          assessment will identify the type of waste (organic,  inorganic,
    ^   etc.)  and general disposition of the  waste (surface, subsurface.
  tamitiat.8°on.  etc.).   In addition,  the potential  for  air  pathway
  36<1.  j.lori by volatile species and particulate  matter  should be as-
   tho8e evelopment of the technical approach will  then  involve  identify-
  ^tei?  activities in the investigation necessary  to characterize the
   e tt ^ tlle "extent" of the contamination and the migration  pathways.
  *• The1Sts  Seneric activities that provide nature- and  extent-type
         t1"""1re-type data pertain to the waste material;  they are some-
             to as "worst-case" data and will  identify  the  chemical
          °f  the waste.   These data,  along with knowledge of  the physical
             waste and the migration pathways  for  the  contaminants  at the
            input into the development of the  extent-type data  collection
   "Potta'    eeP in mind that identification of what is not  in  the waste
      csnt *                                              .    i    •
           in addition to identification and quantification what is

                                 487
-------
                                                                     w
 believed or known to be in the waste.   Therefore, complete chemical c"
 terization of  the waste for every discrete waste area on site is

      The design of the technical approach for the collection of ex
 data  will probably involve many types  of activities, ranging from teC°^
 naissance surveys to the collection of media samples in all directi°D
 from  the waste area(s).   Media samples refer to the air, soil,  surface
 water,  and groundwater around the waste (2).  As such,  the design of yj
 sampling approach to characterize the  extent of contamination is eSge
 for cost-effective site characterization.

                                                                    •X
      The scientific approach to characterizing the extent of contamlP
 involves a series of steps and,  often, multiple sampling events.  Al* ^
 sampling strategies should identify sampling locations  which can be *
 enced to a grid map (1)  that describes the projected maximum range °
 contaminant migration,  including direct surface migration, subsurfac
 lateral  and vertical migration,  surface-water migration, groundwater «('
 contamination  and plume migration,  and transport via air pathway to  $
 soil  and water (2).   Transport modeling can be helpful  in identify*0^'
 extent  of contamination verified by grid-based survey sampling and a ,^
 ses.  Once the borders  of contamination are identified,  grid-based * $
 sampling and plotting can designate the extent of contamination and
 of  elevated concentration.

      Air emissions potential can be assessed using screening and/of  ^yf
 depth technologies as both nature and  extent activities.  Screening  ^
 ities include  real-time instrument  surveys,  head space  samples  of ** ^01'.
 waste,  simple  air monitoring,  and predictive modeling.   In-depth te .$
 ogies include  area emission source  assessment technologies (3)  an<*,^
 emission rate  data from the site in the undisturbed or  disturbed si .$
 (i.e.,  during  mitigation)  conditions.   The need to develop emissi°n .$
 as  part  of  the APA will depend on many factors and these data wil*
 in  the  decision making  through risk assessment.
                                                                   •rX'
      One general  approach applicable to any  stage of the remedial   ^
 gation uses  a  multi-level  sampling  and analytical strategy.   Figurf
 depicts  this strategy.   It  can be employed for sampling  and  analys
 regardless  of  the  stage  of  the investigation.   By employing  screen  ^
 (Level  1),  intermediate  (Level 2),  and detailed (Level 3),  sampl*-0**
 analyses,  data needs  can  be fulfilled  while  using minimum resource
                                                               . .
     Level 1  (screening) sampling  and  analysis  activities  provide    s v
sive survey data which generally indicate  the presence  of  contain^  ..^tft
do not provide species-specific information.  Screening APA can l  ^fg
the potential for air emissions and  help determine  if detailed AP
required.  These data will also help identify the need  to  conduct  &^'
monitoring for worker and neighborhood protection during site nit:  , (r.(
Because the detection is usually performed  on site  and  at  a 1°" c  e
analysis), the survey information  can  be obtained addressing
of samples or site locations.  This  approach is similar to the *
leveled approach described in the  Data Quality  Objectives
Response Activities  (2).
     Level 2, intermediate sampling, involves  fewer  samples  of   ^
yet more analytical information at a higher cost  per analysis.
sampling and analyses generally target  indicator  compounds that_
identified in the nature-type data collection.  Often these  ^  .
pounds are selected due to their mobility which can  help        "

                                   488
-------
    • °^ Contamination.  Level 2 APA may include in-depth area  source
   Sl°n estimate determination (3).
         3»  or detailed testing, involves limited testing with more
     ensive analytical work-up.  These detailed data can be used to
       e Presence (or absence) of compounds identified in earlier  stages
     to support the intermediate level analytical work.
             the proper type of investigative activities  (Table  II) with
          multi-level sampling and analytical approach  (Figure 1) can
      *or a comprehensive and cost-effective remedial investigation.
^hg c  Case study,  prepared for a Superfund site mitigation, demonstrates
8ivg  CePts presented regarding the importance of conducting a comprehen-
ds t   e«ial investigation.  The investigation work is compared against
     e<*ial investigation activities listed in Table II.

                             Site History

 xacpe  emedial investigation and feasibility study were conducted at a
  " it U^er^unc^ site.  At one time, the site was a municipal landfill.
  1 st WaS a P°wer (oil-fired) generating station with surface tanks for
^Vao*. ta^e anc* a waste oil trench.  Later, the site became a metal
   86 yard.

               Remedial Investigation/Feasibility Study
Ik  ^rin
nk  sit     a state ant* EpA hazardous waste site enforcement inspection,
  tyls fi>as Discovered to have moderate levels of polychlorinated bi-
 UtfaC  tP^Bs)  and high levels of heavy metals (lead, copper, zinc) in
            near the abandoned industrial facilities.  Data on the site
       ected in the preliminary assessment and a site inspection; the
       ranked on the NFL.

       r State lead, a total of six stages of sampling and analyses were
        as tl:ie remedial investigation.  The site was divided into three
          areac.   A total of 53 borings were performed and 267 discrete
           were collected (76 surface, 191 subsurface).  About 180 of
lid] ationP    were analyzed.  The sampling program included: surveys using
to  ^ s     ectors and a magnetometer; grid-based surface and subsurface
          k°rin8 and split spoon sampling at fixed sampling depths of 5
          ° to 12 feet« and 15 to ll feet below land surface; and ground-
           g*   No screening APA were performed for volatiles or particu-
          emissions.  The analytical program focused on PCS and metals
               the soilt  The site was evaluated as a site havin8 contam-
           without contaminated groundwater.
 it  ^e f
 I6j. 8 deeaS^bility study used the remedial investigation data from which
                that the remedial alternative would be excavation and
         ln a permitted landfill.   The remedial alternative included a
           °n site health and safety monitoring for particulate matter
        ine monitoring for particulate matter for public protection.

                            Remedial Action

                 12.000 tons of contaminated soils were removed from the
            Process, several physical and chemical hazards were

                                 489
-------
identified that were  not  discovered  in  the remedial  investigation.     ^
III lists these hazards.  The  discovery of hazards negatively affect^
project schedule, the health and  safety program,  and the project bud?

Conclusion

     A significant amount of time, money,  and  effort was spent on cha  ^
terizing the site.  However, the  sampling  and  analysis  plan was bi-&se
terras of the location of  the sampling and  the  types  of  analyses and
incomplete in terms of  types of remedial investigation  activities fe  y/f
formed.  The APA was  essentially  ignored and,  as  a result,  the air *  yp
ing program was developed without complete knowledge of the air cont
tion migration potential.
                                                                  .  r/
     The remedial investigation activities performed are compared 1   ^
IV against the activities listed  in  Table  II.   As can  be concluded_
the listing in Table  IV and the numerous incidences  encountered dur   ^
remediation activities  at this Superfund site,  the remedial investie^
lacked the comprehensiveness necessary  to  ensure  worker protection
performing on-site activities  and to ensure an adequate fenceline 31  .
monitoring plan.  The incomplete  assessment of on-site  contaminants  .
only threatened the adequacy of on-site health and safety measures   .^
public health and safety measures via air  pathway contamination a foj
cal/chemical hazards, but also significantly extended the schedul6   ^v
remediation completion  and resulted  in  significant budget overruns*   .
fortunate that an extremely conservative health and  safety protoco    {j0
followed daily (A); even though personal air monitoring results 1°   ejf
no potential overexposures, personal protective equipment was worn
day, ensuring against overexposure to the  unanticipated contaminsn   ^
discovered during this  remediation project.  These discoveries cou ^^t
seriously affected, for instance, a  remedial alternative with an °n  ->f
treatment where the contaminants could  have adversely affected the
ment technology.
                                                                    c/
                                                               • 4-tl fl   f
     In conclusion, a complete remedial investigation begins witn   ^
hensive design employing many  of the activities identified in   i   <<{
providing nature- and extent-type data and  a multi-level  samplin
ical approach with APA as necessary.  Resources properly  spent  f°*'
complete site remedial investigation can provide  for  a  controlled
action with few liabilities, attainable cleanup schedules,  and  ro1 .e
costs that are within the project budget.   However, despite cojnpr ,^
design and site characterization work, a complex  hazardous  waste
bring new discoveries and unwelcome surprises.
                                   490
-------
     U.S.  Environmental  Protection Agency.  November 1986.   Test
     Methods  for Evaluating Solid Waste.  Volume II:  Field Manual
     Physical/Chemical Methods.   Office of  Solid Waste and Emergency
     Response.
2
     U.S.  Environmental  Protection Agency,  March 1987.  Data Quality
     Objectives  for Remedial Response Activities.   Office  of Emergency
     and Remedial Response  and  Office of  Waste Programs Enforcement,
     Washington,  D.C., EPA-540/G-87/004.

     Clark, J. A.,  C. E.  Schmidt,  T.  D'Avanzo,  "Overview of applicable
     emission measurement technologies for  the measurement of volatile
     hazardous waste emissions,"  EPA/APC  Symposium on Measurement of
     Toxic and Related Air  Pollutants,  May,  1988,  Raleigh,  NC.
4
     Mecham, C.  C.,  J. P. Alexander,  "Worker health and safety/air
     monitoring  case study  of a Superfund remediation project."
     Hazmacon, April, Anaheim,  CA,  1988
                             491
-------
 Table I.  List of Potential Impacts Resulting from
           Incomplete Remedial Investigation
     Improper Type of Treatment Technology Selected

       Ineffectiveness of the Treatment Technology

                Delay of Project Schedule

               Overruns of Project Budget

 Violations of Health and Safety Program On Site via Air
   Pathway Contamination and Physical/Chemical Hazards

Violations of Public Protection Off Site via Air Pathway
 Contamination (Volatile Species and Particulate Matter)

      Nonsupportive of Community Relations Program

       Liabilities Associated with the Remediation
                          492
-------
       TABLE II.  Remedial Investigation Activities That
                  Generate "Nature" and "Extent" Data
 .!!Nature"-Type Data

Historical Records Search

Interviews with Involved
Persons

Media Reports

Enforcement Inspections
and Reports

Limited Sampling of Waste
and Detailed Analyses In-
cluding Air Emission
      s (sample location
   scientific selection)
        "Extent"-Type Data

•  Aerial Survey  (historical and
   current; visible and infrared
   photography, reconnaissance
   reconnaissance observation)

•  On-site Remote Sensing
   (geophysical techniques)

•  On-site Transport Modeling

•  Screening Surveys Using Real
   Time Analyzers for Volatile
   and Particulate Matter
   Emissions

•  Survey Sampling and Analysis
   of Waste and/or Media

•  Grid-Based Sampling for
   Lateral Extent and Intermedi-
   ate Analyses for Indicator
   Compounds of Waste and/or
   Media

•  Grid-Based Sampling for
   Vertical Extent and Inter-
   mediate Analyses for Indicator
   Compounds of Waste and/or
   Media

•  In-Depth Assessment of Air
   Emissions Potential

•  Contour Plotting and Cross
   Sections
                           493
-------
Table III.   Listing of Physical and Chemical Hazards
            Discovered in the Remedial Action Not
            Identified in the Remedial Investigation
     Three Unlabeled,  Buried,  Compressed Gas Cylinders

     Unlabeled,  Buried 55-Gallon Drum Filled with
     Unused Waste Oil

     Asbestos Found in Two Locations:
     1) Approximately  20'xl5'x2' thick at a depth of
     1 to 3 feet below land surface; and
     2) Small area 10  to 14 feet below land surface

     Polynuclear Aromatic Hydrocabon (PAH)  Material
     26 to 27 Feet Below Land Surface in One Area

     Surface Soils Contaminated with Pesticides
                        494
-------
      Table IV.  Comparison of Remedial Investigation Activities
                 Performed to Those Activities Listed in Table II
               Gathering Activities
                                                    Site
                                           Remedial Investigation
   lstorical Records Search
       Preliminary Assessment Activities
1  L*
    ited Sampling of Waste and Detailed
                                         • Completed

                                         • Not Reported

                                         • Analyses Limited to
                                             suspected contaminants,
                                             not comprehensive .
                                             No screening APA were
                                             performed.

1 — •£&>e_J)ata Gathering Activiti
                               ties
            Site
   Remedial Investigation
        Survey

  SC
     lte Remote Sensing
       e Transport Modeling
           Surveys Using
             Analyzers
   to
   Waste
of
                     Analysis
            and/or Media
      B
            SamPlin8 for Lateral
          and Intermediate Analyses
     --
    «ts     SamPling for Vertical
          and Intermediate Analyses
     g
     tc,   Assessment of Air Emissions
         P1
         rj-otting and Cross-Sections
• Not Reported

• Magnetometer Survey
    but it was confused due
    to surface scrap metal

• Not Conducted

• Radioactive Material Survey
    Volatiles Species Surveys
    Not Performed

• Extensive Survey Sampling
    with Limited Analytical
    Work

• Completed
                                           Sampling Biased to Surface
                                             and Five Foot Spacing In-
                                             ternal Sampling. Sampling
                                             should have been continu-
                                             ous to groundwater .
                                             Analyses limited to
                                             suspected compounds
                                             with no provision for
                                             detailed or comprehensive
                                             rehensive analyses.

                                            In-Depth APA Not Performed
                                           Incomplete.
           Sampling Plan.
                                 495
-------
   Sampling  Activities
Analytical Activities
                            LEVEL  1
   Survey Measurements,
     Screening APA

   Grab  Samples  (Limited) of
    Waste Material
     (Worst-Case)
Real-Time Detection
Detailed Analyses
                            LEVEL 2
  High Number of Samples
    Waste and All Media
    (soil, waters, air),
    Detailed APA - Air
    Emission Estimates
Intermediate Level of
  Analyses, Targeted or
  Indicated Analytes
                            LEVEL 3
  Limited Number of
    Samples, Waste and
    Media (soil, waters,
    air)
Detailed Analyses and/
  or Confirmational
  Analyses
Figure 1.  General Approach for the Sampling and Analytical Stra
                                                                tegf
                                   496
-------
fOXlr *EST EVALUATI°N  OF A METHODOLOGY FOR MEASURING EMISSIONS OF SELECTED
  1C METALS FROM  STATIONARY SOURCES
Gl
Kadi" °* Osn>ond, Winton  Kelly
Res.6n c°rporation
   arch Triangle  Park,  North Carolina 27709
   ja-  _
ti. s   e- Ward, Thomas  Logan,  M.  Rodney Midgett
^SL 'B!rnvironn>ental  Protection Agency
ReSe"RTp/QAD/Source  Branch  (MD-77A)
    rch Triangle  Park,  North  Carolina 27711
       the request  of  the  U.  S.  Environmental Protection Agency (EPA), a
8aniplJtaslt research  effort  has been implemented to develop a validated
6mS6.n8 find analytical technique to measure multiple metals in the
1easur°ns ^rom  stationary sources.   The methodology was designed to
C°Ppe * ^e f°H°winB 16 toxic metals:   lead,  zinc, phosphorous, chromium,
       nickel> manganese,  cadmium,  selenium,  arsenic, mercury, beryllium,
      a> silver, antimony,  and barium.

    .  sampling method was based upon extensive literature and
       ^ studies •   The results  of these studies indicate that the most
        desi8n of  Che sampling  train  is a modified EPA Method 5 train
         its  Particulate  collection efficiency, ease of operation,
  ttet i   v' and cost-  The absorbing  solutions identified to collect the
   an  * Deluded nitric acid, hydrogen peroxide, and acidified potassium
     *an*te .   The configuration  and components of the sampling train
   er vd an EPA Method 5 glass probe with glass probe tip, a heated
   «ct<   containing a quartz fiber filter, an empty condensate-
P*toXid   lmPlnger,  two 5  percent nitric acid/10 percent hydrogen
S     e lingers,  one implnger  containing acidified permanganate, a
         impinger,  and the usual EPA  Method 5 meter box and vacuum pump.
         an extensive  laboratory study,  the results of which are
     n   ln a Paper  by  Cole  et al.  in this symposium, a field test
  CO*BB« Was dev*loped and later performed to determine the ability of the
 i  'Ion   d 8ainPling  train to collect the 16 toxic metals from a
 %*10 aty source.  The  source for this field test program was the
 ,  the    of * municipal solid waste incinerator.   After metal collection
 %tn '^PUng  cratn as a whole was studied, the  back-half impingers were
 t  coii "" flrst  to  see lf the metals had reached them, then to determine
     ni* °tlon characteristics of the five -impinger arrangement.  The
     ti  *l test  approach was formulated to compare the relative
         efficiencies  of the recommended sampling train and an alternate
        vtaln usln£  the same five impinger configuration, but with a
       absorbent  strength (i.e., 0.1 N HNO. instead of 5 percent HN03) in
  tCu*v   ln)Pingers.  Furthermore,  samples were collected to compare the
   HetvC°llection efficiency of the proposed sampling train to that of
      **od 1 ni * £
         1U1-A for  mercury.

  l    * tesults of the  analytical data analyses indicated no significant
         betweftn the metals collection ability of 0.1 N nitric acid and
         nitric acid.  The recommended sampling train was also found to
       8tically equivalent to the EPA Method 101-A in collecting mercury.
                                  497
-------
                                                                     ill*
Furthermore, front- and back-half metal distributions indicate that, Wi
the exception of mercury, arsenic, barium, and phosphorous, most of c^e
metals were captured in the front-half or filter section of the train-

Introduction

    The U. S. Environmental Protection Agency (EPA) is considering    j
regulating toxic metals emissions from incineration processes because  i
the potential environmental and health impacts.  Toxic metals are rel*
to the environment in stack effluent when industrial incinerators buffl,s|i
materials containing trace metal contaminants.  The potential health f
has prompted EPA to develop and validate a methodology to quantify   ^J
accurately the emissions of 16 toxic metals in stack gases from staCi0
sources.  The toxic metals of interest are as follows:  lead, zinc,
phosphorous, chromium, copper, nickel, manganese, cadmium, selenium,
arsenic, mercury, beryllium, thallium, silver, antimony, and barium-

     Radian Corporation under contract to EPA's Environmental Monit°r
Systems Laboratory (EMSL) has performed an extensive research effort
develop a diverse, multi-metals sampling methodology.  The technical
approach used to formulate this methodology had three phases:  (1) * 0{
literature review, (2) laboratory studies, and (3) a field evaluati0** ^
the sampling protocol.  The literature review was undertaken to dete  ,
the source of toxic metals emissions in stationary source effluents. ,t
to identify potential sampling and analytical techniques for these J ^
metals emissions.  Based upon the literature review, a sampling era*
absorbing solutions were recommended as well as an optimal analytic*
method.  Laboratory studies were then conducted to determine overal1
precision and accuracy of the analytical portion of the methodology  ^
including sample preparation and analysis.  The final phase involve
field evaluation of the sampling protocols.

     The overall objective of the three-phase research effort was c  $
                                                                 *tf&
develop an analytical and sampling methodology that is both accurav ^
repeatable.  Specifically, the methodology was designed to accommoo   «i
both a wide variety of sampling conditions and sampling concentrat   ^(>
a form conventional enough to be adopted by the sampling community-  ^J,
report presents the procedures used to conduct the field evaluati°n  fit
and the results of the 16 metals analyses performed on the r«ilfteC
samples.

Experimental Approach

     The primary objective of the field test program was to
proposed five-impinger design and reagent strengths on the coll«c
the 16 trace metals of interest.  The technical approach used in
development of the field study was designed to address each of tn
objectives listed below:

          Determine if impinger solution strength has a quantif*-8  ^n*1"
          effect on the amount of target metal collection in th«      ^
          impinger solutions.                                   , 0( &*
          Quantify for each target metal the location (front hal
          half) in the sampling train where each is predominately
          collected.
                                   498
-------
         Compare  the  amount  of mercury collected by the  proposed
         multi -metal  train to  the  amount of mercury collected by the EPA
         reference  method  for  mercury  (EPA Method 101-A) .
         Evaluate the potential of manganese contamination from the
         permanganate solution.

      *  ^    test  program was conducted at a municipal  solid waste (HSU)
   st v   assumed to contain  sufficient  quantities  of the  16  metals  in
     ac* effluent to meet analytical detection  limits.

       Matrix

       Satnpling matrix was designed to evaluate  the four objectives
te«t c previously.  To achieve  the desired information,  the  following  two
     ttParisons were formulated,

                          Test Comparison #1

    Ru° C0nsecutive comparisons were made  of the  five- imp inger
    80fatlon (Cl) to an identical sampling train  (Ml)  that  uses  impinger
 *s to utions of lower concentrations.  The purpose  of this comparison
c°llect?Valuate the effect that impinger solution  strength had on the
         of th-e 16 targeted metals.  Train Ml was designed with the
ja          -                     .
:B|Pti1geent configuration as Cl with  the exception of the concentrations of
 •1 N n.s 2 a"d 3.  Train Mi's impingers 2 and 3 were designed to contain
a^ train *ic acid/10 percent hydrogen peroxide, whereas impingers 2 and 3
       C1 contained 5 percent nitric acid/10 percent hydrogen peroxide.

                          Test Comparison #2
   TV&
    u  c°nsecutive comparisons were made of the five -impinger
      atl°n (Cl) to that of the EPA reference Method 101-A (E101) for
        The purpose of this experiment was to compare the amount of
     io?llected by sampling train Cl to that amount collected by EPA
    ci       With ttlts comparison, the relative accuracy of sampling
       to EPA Method 101-A can be determined.

   a«t*?^ln8 ^e two comparisons above, the amount and distribution of
   Hi v   in the acid impingers (impingers 2 and 3) of sampling trains Cl
       ke evaluflted for trends.   Front-half (filter and probe rinses)
»ra   lf ^tne Method 5 impingers) metal splits were determined for
^c°Hstt t° throu6n the two comparisons.   In addition, because manganese is
t % 0£Uetlt of the potassium permanganate impinger (impinger 4), as well
^ to vhe(.Lhe 16 netals of interest, a qualitative determination was made
  p^tiB      forward migration of manganese occurred during this field
     * Program.
Xti
    l
         D
         Results  and Conclusions
     * n
     n  r°p°sed sampling train using 5 percent nitric acid/10 percent
V I8urat?r°xide  was  tested and compared with the same train
^ J tuns     using 0.1  N nitric acid/10 percent hydrogen peroxide.  Two
    c°Cre Perfornied in which two samples were collected using each
                  simultaneously.
                                 499
-------
                        Absorbing  Solution  Strength

     There were  five metals detected  in  the acidified hydrogen
 fraction  on which  a comparison  could  be  made:   lead, zinc, barium,
 manganese, and mercury.  A two-way analysis of  variance  showed no
 statistically significant difference  in  the amount of these  five
 collected in the different-strength acid solutions.  Therefore, at tb* t
 statistical levels tested, the  acidic hydrogen  peroxide  solution str*1*
 was not a factor in the metals  collection efficiency.

                            Metals Distribution
                                                                      .(I
     For  the combination of metals detected at  this municipal solid tf*u
 incineration facility, the following  conclusions were made concerning
 part of the sampling train where the  metals were collected.
                                                                       tj
     The  following metals were  primarily and consistently collected *
 front half (probe catch and filter) of train Cl:

        Metfll                                Percent in  the  Front,

       Cadmium                                           100
       Lead                                              98+
       Chromium                                          100
       Nickel                                            81+
       Copper                                            100
       Zinc                                              98+
       Antimony                                 64-100 (mixed results'
       Selenium                                          100

     The  following metals were  collected primarily in the back
 (Impinger fraction) of train Cl:

                                              Percent in the
       Mercury                                           98+
       Arsenic                                          100
       Barium                                           100
       Phosphorous                               85+ (mixed
       Manganese                                        100

Beryllium, silver, and thallium were not detected in sufficient 4U
during this test to predict where they would be collected in trai-11
general, train Ml collections were similar to the above.  These f
were consistent with those of previous studies for lead, cadmium.
and chromium in which essentially all of these metals were coll*c   ^«
the front half of the Method 5-type train.  The results for mercuy^f*
arsenic were also consistent with known volatility for mercury a°
arsenic materials.

                       Mercury Collection Efficiency                  .$
                                                                 otit
     The proposed sampling train (Cl) included a final impinger c     ,
acidified potassium permanganate for the collection of mercury*   -je^
efficiency of this configuration was unknown, and tests were p«*£
using EPA Method 101-A as a reference.
                                   500
-------
    c * r«sults of these tests showed that the proposed sampling  train
8tatistf       mercury as efficiently as the EPA reference procedure.  The
       *1 analysis of variance showed no significant difference  between
       ts of mercury collected by the proposed procedure, the modified
    d i procedure that used a weaker acid absorbing solution, and the EPA
    9o        An unexpected finding was that the largest fraction, better
S*ctlon percent, of the collected mercury was in the acidic impinger
      °f the proposed sampling train.

ctiiiipa **^ °n these test results, the proposed sampling train (Cl) is
C°llecti   staclstically to the existing EPA Method 101-A for mercury
      ?n>  However,  because of analytical requirements in analyzing for
         a ny°*r°Een peroxide matrix, there may be circumstances  where
      °d 101-A- may be desired instead of the Cl train.

                        Manganese Contamination

^* tyaj*n °^ t*le eight Cl train results showed low manganese results.
 ° contto«*lad Sn obvlous^y high manganese content due in all probability
 **d f0  Cation from the permanganate solution.  This demonstrates the
 ' heated Blass  Probe and high purity quartz fiber filter,
lh  fl«d  an imPinBer  train containing acidic hydrogen peroxide and
lh l* n,fttpermanganate ,  may be used to collect and quantify, in general,
'»! *tf*n u  °f lnt«rest.  Because the field test program indicated that
   l°n a   Of che acidified hydrogen peroxide is not a factor, either
          be used.
      n»ftivld study has provided preliminary estimates of the precision
      ar     for determining 13 of the  16  metals.   Additional field
    o*> * necessary  to establish final  estimates of precision.  In
Hj^tt' !!° Precision estimates  could be made for three of the metals of
  * for I-L thls source.   Further field tests would be required to collect
      thftse metals.

lnttl»« hy$° the analytical  requirement associated with measuring mercury
       to8en peroxide matrix and the potential for manganese
        e from the  permanganate impinger  solution, mercury may be
                                 501
-------
quantified In a separate sampling train (EPA Method 101-A).  In such
cases, the acidified permanganate impinger would be deleted from the
proposed train and no mercury analysis would be performed on the Cl c

Recommendation for Future Research

     The field sampling program was intended to provide guidance i°  «
selecting an appropriate proposed method for measurement of toxic n>e  $
emissions.  Further research is needed to fully validate the procedu
all 16 metals and to provide better estimates of the method precis*0
These research areas are described below.

                            Collaborative Tests

     A full-scale collaborative test using sampling techniques
those used in these studies should be performed to establish bette*
estimates of precision and to further test the proposed method.

                 Tests of the Method Versus EPA Method 108

     The sampling procedure has been indirectly tested for its ab*1 tj &
collect certain volatile metals (As, Hg, etc.).  Further specific e ^
this method as compared to EPA Method 108 (specifically for arseni0'
verify the adequacy of the sampling procedure.

               Modification of the Acidic Impinger Solution
                                                                  J f»  C
     The acidic hydrogen peroxide solution was used in the propos*  flo>
for two reasons.   First, an acidic medium tends to promote dlssolu  ^
metal salts and therefore aids retention in the impingers.  Secofl i  ^
hydrogen peroxide is needed to protect the succeeding impinger S°]J
from oxidation by SO,.  However, if the permanganate solution 1s
from the sampling train, then hydrogen peroxide Is no longer mce
its primary purpose.  Tests should be performed to determine if
peroxide promotes collection of the other volatile metals.  If n°
could be eliminated.

DISCLAIMER

     This paper has been reviewed in accordance with the U. S.
Environmental Protection Agency's peer review and administrative
policies and approved for presentation and publication.
                                   502
-------
   ACK PM1Q  SAMPLING METHODS:  A REVIEW OF BASIC REQUIREMENTS
      E;
       Research Institute
       m,  AL 35255
   , g ental Monitoring Systems Laboratory
 eseav. f^ronmental Protection Agency
    ch Triangle  Park,  NC 27711
  e a t metnods  have been developed for measuring in-stack PM10,  particu-
^at em i r °^ nominally 10-/im aerodynamic diameter and smaller.  Devices
?86(1 In lTy particle inertia for size fractionation and duct traversing are
K  t*Um«,   ^ approaches.  The emission gas recycle (EGR)  approach uses new
      ntation to eliminate the conflict between isokinetic sampling and
          of constant flow rate through the inertial device (to maintain
          e separation).   The simulated Method 5 (SIM-5) approach uses the
        °d 5 or 17 sampling train and a specified sampling protocol which
           due  to anisokinetic sampling to within acceptable limits.  Both
         require new specifications on the geometry of sampling nozzles.
       s and disadvantages of both approaches are discussed.
   M
   lotiSUr6lBent:  of PMi°  emissions from stationary sources requires the
   such   SlzlnS capability to sampling techniques that have been widely
           for  Methods  5 or 17.   Because the measured emission rates are
          relate to  health effects of measured ambient concentrations, the
      O   size  is the more important parameter rather than the physical
   st21  er related  parameters.   Two types of instruments that provide
   ea an8 capability and have been widely used to characterize control
    Of ** inertial impactors and cyclones used in sampling trains like
   tencettl0ds  5 and 17>  although with significant operational
        8 from  Method 5 and 17 procedures.

fJ* «f t,°Verall  quality  of the measurement depends upon several factors.
*»J? v**t Se factors.  spatial variation of the emission rate (stratification
*tS ation    of both the Bas vel°city and concentration) and nozzle
5« ^tat-?065 f lcient (anisokinetic sampling errors),  relate to selecting a
6tn<1 l? b    sample.  These types of error are readily limited in Methods
Of % ^u J[ sampling at the center of multiple zones across the sample plane
Vi  **»  traversing)  and by adjusting the sample flow rate at each point
Hltlo«al6rSe to match  the nozzle and stream velocities within ±10%.
Villfttic  size  measurements conflict with these requirements for
*OUcli viSampling and traversing, because aerodynamic size cut would vary
'O"8- u i   the needed  flow rflte variation encountered in most process
t» tjtses     s the lnlet diameter of the sampling nozzle were varied during
V  W     most desirable approach identified to resolve this conflict
*t *ftte ^endent control of the  sample flow rate from the duct and the
Sui^que   Ugh the  size separator by recycling cleaned, dried sample gas,
Si  s a  °alled Emission Gas Recycle (EGR).1-2  Because recycle flow
\j°*5 /SJ*W sampling system,  an alternative approach, called Simulated
\1 fitit CC  5)-  was  also identified for development.   It requires minimal
*V^ati   ges and  ls  raore lik« traditional size measurements.  After
   **par °" °f  net error from stratification, anisokinetic sampling, and
          l°n»  a practical approach was chosen in which the requirement
                                 503
-------
 for Isokinetic sampling is relaxed, the number of traverse points is
 reduced,  and the dwell time at each traverse point is proportional to
 velocity.3•*

      The  PM10  techniques summarized here address measurement requirements
 for emissions  in the particulate form at stack conditions, before the
 emissions are  exhausted into the atmosphere.  All testing has been performed
 with the  filter in the process stream.   However, the use of an out-of-stack
 filter or other approaches for measurement of particulate emissions that are
 in  the vapor phase at stack conditions  is not precluded.  This paper gives i
 review of the  basic requirements of these in-stack PM10  techniques for
 widespread  utilization.


 In  Situ Size Separation

     Although  the  filter can be in or out of the process duct,  the size
 separation  device  must be located at the front end of the probe to avoid
 errors caused  by turbulent deposition (or turbulent diffusion)  and settling
 of  particles in the probe.   Comprehensive models,  based  upon extensive
 experimental data,  are available for each of these mechanisms.5-6   These
 models indicate that deposition in probes of practical dimensions  can be
 minimized to about 5 and 25%,  respectively,  for 5-fan and 10-^m particles, by
 selection of flow  rate.   However,  actual deposition is predicted to be
 typically higher because minimum deposition  occurs over  a narrow range of
 flow rate, which generally does not correspond with the  appropriate flow
 rate for  any given size  separator.   Such probe losses indicate  not only that
 the size  separator should be at the front end of the probe but  also that the
 probe  should be washed if the  filter is  not  immediately  behind  the size
 separator.
EGR Principle of Operation

     A block diagram of the EGR train  is shown  in  Figure  1.   Stack gas  is
isokinetically extracted through the sample portion of  the  EGR nozzle,  where
it combines with recycled process gas  to provide a constant flow rate
entering the sizing component of the sampling train.  After passing the
inertial sizing device(s) and in-stack filter,  the combined sample and
recycled gases pass through the heated probe, condenser,  and dry sorbent or
impinger train, and into the EGR flow  control module.   As in conventional
Method 5 control modules, this gas flow rate is controlled  by coarse and
fine control valves (Vx and V2) at the entrance of the  sealed pump.  At the
exit of the pump and absolute filter,  the total flow rate is measured with a
laminar flow element (LFE).  The gas stream is  then split into the recycle
and sample flow lines.  The sample flow is monitored in the normal manner
with a dry gas meter and a calibrated  orifice.  The partitioning between
sample and recycle gas is controlled by valves  V3  and VA , located downstrean
of a second LFE.  The recycle gas line, along with the  sample and pitot
lines, passes through the heated probe in which the recirculated gas is
reheated to the duct temperature before entering the sizing component with
the sample stream.

     Operation of the EGR train is similar to standard  Method 5 or 17
sampling.  Selection of traverse points and sample flow rate is the same as
for Method 5 or 17.  In practice, Vx and V2 are first set to regulate the
total flow rate to create the desired  size cut; then the  sample flow rate is
adjusted to be isokinetic by using V3  and VA at the first traverse point.
Changes in the recycle flow rate alter the total flow only  slightly, so that
only one additional adjustment of V2 is usually needed  at each traverse

                                    504
-------
  * ^fusin? !        adjustments of V4 .  The only aspect  that may  initially
  8te is ad?*  V? exPerlenced Method 5 operator is  that the  sample flow
 MVes<  °Justed by using the recycle valves rather  than the  total  flow
        A mn^ complete description of the EGR technique  is  available,* and
               operation and maintenance manual is  currently in review.
               of Operation
    To
 ? Sl*l-5r^;e *inimal changes in traditional sampling systems in developing
 N*menr     "'  ^ approach was Allowed that minimizes expected
  ^Izes    error and ensures it is within known limits.   This procedure
 .>Un» .SamPling efforts while keeping errors caused bv anisoklnftf^
                                is within  known  limits.   This  procedure
                 eff°rtS whlle  keePing errors caused by  anisokinetic
4,Ss*ons °mPa5able to those caused by spatial and  temporal variations  of
o* UU duct ^nisokinetic sampling bias is kept in this range by synthesizing
V&i^erent rfaVerS6    m partial traverses, if necessary, by using nozzles
(i cl*y)  »4?uainet:erS t0 keep the velocity ratio, R  (- duct velocity/nozzle
th6'' each       acceptable limits.  Points for  each partial traverse
«* ^ohibfn°Zzle slze)  are selected that have duct velocities  in a range
*h dl*ig +?nf err°r dUe  t0 anlsokinetic sampling of 10-^m particles from
'N Otie no, i  a5 eaCh P°int'*   Tne resulting limits are broad enough so
»tid  °an h!       usuaHy sufficient for an entire traverse.   Actual net
C ^gativ! expected to  be much I688 because of cancellation of positive
ls?6ase WithSS?P   S errors from P°int to point.   Furthermore,  these errors
O of that J   ?2Uare  °f the particle  size (the error for l-Mm particles
V8veU hpl    ii  Pm particles)'  flnd PM10  includes  substantial mass at
Hn tllan in   W    tim'   F°r most sources,  the  particulate  mass  at sizes
""- inB to  /m    sufficient to  limit the  error due to  anisokinetic
  '6ntUl of In  than  ±11%  Wlth the  SIM"5 Protoc°l-*   As  stated  above,  this
  h?alUve   ~i    error would rarely be encountered  because of cancellation
    n8 Plan  negative  s^Piing errors incurred  from point to  point  in the

n». ^
                                     and v*--  for a given nozzle and
                           from closed form equations of nQl/2  and u,
    " ts th      eermne   rom cosed form equations of nQl/2 and u,
    sited  ? Process gas viscosity (in micropoise), Q is the flow rate  for
   1          cut(s) in the inertial sampler, and u is the nozzle
       re!                                e,  an  u  s   e nozze
  iS for ^UltlnS from  its  inlet  diameter and the  required flow rate.
    *£  ±2o»n°2Zle are  broad  at low  duct  velocities  and decrease to minima
       jt r   St the highest  velocities.   For a  typical combustion gas  and'
                                         Vmin  a"d V-nar  «6  0.62u and
                  at a nozzle velocity of 10 m/s,
    j                                   •
     tt^637 of req"ired traverse points affects both sampling error and
     totaf C°St to Perform measurements.   The  error due to  spatial varia-
    Vfts a  Articulate emissions as a function of the number of  traverse
    0utcesna    d by Shl6ehara-7  His results for actual field data from
'V * one  ndicate that 95% confidence intervals decrease rapidly from
Vj '  It fsamplinfi P°int to less than ±10% at eight or more traverse
Nt?Ul*t* ^.exPected that,  in general,  PM10  is less stratified than  total
\J tcati        S'   However.  the limited data available indicate that
 It**7  t   °f PMl°  ls  not  substantially  less than that described by
    * ^an^in01181  Particulate  matter. 3  Thus,  to limit this type of error
    nfc ait    *'  elBht °r  nine  traverse  points are needed at the optimum
    86 Pol68  described in  Method 1.   For other sampling locations, 12
      *-«ints are required.
      ** dlfference between  the  SIM-5  protocol  and Method 5 or 17 is the
    6 Poi    Methods 5 and  17,  the  dwell  time  is the same  for all
        nts-  The measured concentration is a velocity-weighted average
                               505
-------
                                                                    new
for all points, as it should be for determination of emission rate,
the sampling rate is varied from point to point proportional to the  ,
velocity.  Because flow rate cannot be adjusted with a SIM-5 PM10 s  jj
the dwell time at each point must be proportional to the stream vel°c
obtain a velocity-weighted sample.
Inertial Samplers

     A single stage inertial size separator offers the advantage of e^
determination of PM10 compared with multiple stage devices.  For s*°* $
stage sampling, cyclones are recommended because of their immunity
of the difficulties with impactors, such as particle bounce, anomal0
reactions between process gas and collection substrates, and overlo
However, if the size distribution of the particulate emissions is ""
conjunction with determination of PM10, then cascade impactors are
the best approach, subject to precautions to avoid potentially hig"
because of the difficulties described  above.  Although series cy-1"
provide multiple size fractions, the mass loading necessary for
sample retrieval requires unacceptably long run times for the outle ^
trations of most exhaust streams.  With cascade impactors, multipl6
with nearly the same size cut can be utilized to avoid the effects   .
particle bounce and overloading, and preconditioning of substrates
runs can be used to avoid the effects  of substrate reactions.8
     Techniques for PM10 have been developed and evaluated by em
Cyclone I of the EPA/SRI Five-Stage Series Cyclones8 for the size
Manuals giving details for the EGR and SIM-5 procedures for Cyclop
performance specifications are in the review stage.  An EGR nozz *L
been developed for cascade impactors.  In the SIM-5 manual, Proce fl0t
performance specifications for cascade impactors are given to aug"11
impactor manuals.


Nozzle Requirements                                                   J

     It is well known that the severity  of particulate deposition ^0^1'(
hook or bent nozzles with constant inside diameters renders these &#i$
unacceptable for use in PM10 measurements.  Common geometries f°  ^ p
sampling nozzles have also been  found to cause  significant shifts^e
cut of Cyclone I and significant collection of  PM10 particles in   0{ ,
nozzle.10  Thus, verifications and specifications for the geome* *tf y
nozzles beyond those required by Methods 5 and  17 have been nece ^p
PM10 emission measurements to limit errors associated with tnesedo po'y
problems.  For single stage  cyclones, nozzles are required that   (t^cj1
affect the size cut of the cyclone and that do  not collect PM10 ?ofe** (l
above a specified limit of collection efficiency,  Impactors are ^ ^
conditions that take losses  and  effects  of the  nozzles on the s   ^
the first stage into account, and laboratory calibration data a
the basis for determination  of the operating conditions.              p
                                                            .  , to <>
     For EGR measurements, two recycle nozzles  are recommended  ^ o
range of duct velocities  from approximately 15  to  110  fps. Eac   $o*e y,
has a velocity range of more than ±60%.  For SIM-5 measurements,  ^j
nozzles are necessary.  A recommended set of SIM-5 nozzles for ^^ ^
operating conditions relevant to evaluating its capability are   ^te
in Table I.  For typical  combustion gas  at 300  DF, the PM10  fl  ^  $
Cyclone I is 0.6 acfm.  This flow rate prescribes  the  limits o
velocity for each nozzle, as shown  in Table I.   The values ^T
maximum velocity variation refer to the  percent variation of
                                    506
-------
        below its  average that can be accommodated with one nozzle,
         fc^e  capability for performing a complete traverse without
  ):a    the nozzle.   The maximum variation can be accommodated when the
   c   tCt vel°city is  the same as the nozzle velocity;  the minimum varia-
  eeu  °e accommodated  when the average duct velocity falls between two
    e   e nozzles  of the set.   The use of more closely spaced nozzle inlet
      8 increases  the minimum  variation that can be accommodated.

    Datn „>
     Ql-a (Precision and Relative Accuracy)
       field  studies have  been performed to develop and characterize the
     .* methods using  Cyclone  I.   As  measured by a dual probe technique,
 -.tuu "*sif\

^ tonnereSearch reported in this article was supported by  the U.  S.
  tact n  ^ Prt>tection Agency under Technical Directives 5,  14, and  21  of
      n°- 68-02-4442.  Thomas E. Ward was the Task Officer.
  tnPap6r has been revlewed in accordance with the U. S. Environmental
      f A8ency's peer review and administrative review policies and
       °r Presentation and publication.
      p.: B- Harris, L. Beddingfield, "Isokinetic Sampling with a  Fixed
       *°w Rate Device Using Exhaust Gas Recirculation", Presented at the
        ird EPA Symposium on Advances in Particulate Sampling and
        asurement, Daytona Beach, FL (1981).

       ' D> Williamson, R. S. Martin, D. B. Harris, T. E. Ward, "Design
        « characterization of an isokinetic sampling train for  particle
      (198 measurements using emission gas recycle", JAPCA 12: 249-253


                                 507
-------
  3.
  4.
  5.


  6.


  7.
  8.
    W. E. Farthing, "Evaluation and Recommendations of Protocols f° „
    PMir in Process Streams: Recommended Methods, Volume I,"   rrefl
    in draft form as SRI-EAS-83-1038, under EPA contract no. 68-02'
    3118, Southern Research Institute, Birmingham, AL.

    W. E. Farthing, A. D. Williamson, J. D. McCain, T. E. Ward,
    "Evaluation of Protocols for Size-Specific Emission Measures
    paper 85-14.3, In proceedings of the 78th Annual Meeting of
    Detroit, MI (1985).
N. A. Fuchs, The Mechanics of Aerosols. McMillan Co., New
(1964).

B. Y. H. Liu, T. A. Ilori, "Experimental observation of
deposition in turbulent flow," J. Aerosol Sci. .  £: 135-145
                                                                 o
                                                               o
                                                               (1?
    R.  T. Shigehara, "Proposed Revisions to Reduce Number of      ,
    Points in Method 1 - Background Information Document, EPA-450/ ,
    82-016a, U. S. Environmental Protection Agency, Research
    Park, NC (1982).

    J.  D. McCain, S. S. Dawes, J. W. Ragland, A. D. Williamson,
    Procedures Manual for the Recommended ARB Particle Size
    Distribution Method (Cascade Inroactors). California Air Resou
    Board, NTIS No. PB86-218666/WEP.

9.  W.  B. Smith, D. B. Harris, R. R. Wilson, Jr., "A five-stage
    system for in-situ sampling," Environ.  Sci. and . Technol^.
    1387-1392 (1979).

10. A.  D. Williamson, W, E. Farthing, T. E. Ward, M. R. Midgett,
    "Effects of Sampling Nozzles on Particle Collection Charact*
      of Inertial Sizing Devices," In: Proceedings of the 80th
      Meeting of APCA, New York, NY (1987).

  11. S. P. Belyaev, L. M. Levin, "Techniques for collection of
      representative aerosol samples", J. Aerosol Sci, . 1(4): 325
        Table I.  SIM-5 nozzle diameters and velocity limits.


Nozzle
Diameter
(in.)
0.136
0.150
0.164
0.180
0.197
0.215
0.233
0.264
0.300
0.342
0.390


Nozzle
Velocity
ffps)
101
83
69
58
48
40
34
27
21
16
12
Minimum
Stream
Velocity
A-0 . 8*
(fps)
76
62
50
40
32
24
18
13
10
8
6
Maximum
Stream
Velocity
A-l . 2*
(fos)
124
103
87
73
62
53
46
37
31
24
18
Stream
Velocity
Between
Limits
ffosl
108
90
74
62
51
42
35
28
22
17
13

Minimum
Velocity
Variation
(±l)
15
15
17
19
21
25
31
35
39
40
40
*The aspiration coefficient for 10-jim particles, A, is that g
 Belyaev and Levin.11
                                   508
                                                                25
                                                                26
                                                                29
                                                                31
                                                                35
                                                                41
                                                                45
                                                                49
                                                                50
                                                                50
-------
                                                       EGft PROBE ASSEMBL Y
O7
O
CO
                RECYCLE
                LINE

                                                                            I	      I
                                                          SEALED PUMP
                                                                                                                              EXHAUST
                                                                                                  DRY GAS METER
                                       Figure J. Schematic of the emission gas recycle (EGR) train (Williamson etal?).
-------
DEVELOPMENT OF METHODOLOGY TO
MEASURE CONDENSABLE EMISSIONS
FROM STATIONARY SOURCES
J.D. McCain and A. D. Williamson
Southern Research Institute
Birmingham, AL 35255

T. E. Ward
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
     The basis of the standard for ambient air particulate concent* ^
was changed from total particulate concentration to PMj„ concentr* f0(ft>
This change may result in the need for updating source inventory in  jo.
tion to include PMj 0 emissions.  Because condensable matter is l^^jjafl ,
represent a greater fraction of primary emissions in terms of PMiO  $ t&'j(
the case for total emissions a source test method should be develop  ^^
includes the condensables component of the emissions.  A review °   op*1
techniques was conducted from which recommendations are made fnf
mum approach to developing such a method.
                                   510
-------
        DEVELOPMENT OF METHODOLOGY TO MEASURE CONDENSABLE EMISSIONS
                         PROM STATIONARY SOURCES
thfee nt.°f  "Primary" particulate emissions from sources and the third is
I"*?98 tor    °f "secondary"  aerosols.  Sulfates formed from SO2 and organic
, ilttary em*d  fcom Phot°chemical processes are examples of secondary aerosols.
i  lia or"1!1*310118 are coraP°sed of t1)  materials that are in particulate form
 1  later •  quid) in the gas  streara before discharge to the atmosphere and
       £    that *re ln the vapor Phase before discharge but that rapidly
       -  8dd  t0 the Particulate burden.   This condensation takes place
i     ai1*9  aS  the materials mix wlth ambient air after discharge.  Since
 ttet       takes Place preferentially on small particles, condensable
           entrated Priraarilv in the ^10  fraction of source emissions.
           data on condensable emissions may be needed to set up State
           n Plans for compliance with ambient air PMj 0 standards.

Of1 8tatior°blem °f dealin9 witn  condensable matter  in emission regulations
«   * Meth   y sources nas Ion9 been a difficult one.   During the development
fto'^its    5f the reference method  for measuring  particulate emissions,
c ri**l aimsj re adyanced  for specifying a sample collection temperature in the
lq'ilj«tat     range ^  that condensable matter  would  be  included.   Practical
l"c thfi fin1?"8' esPecially of  filter wetting  due to  condensed moisture, led
iith^°Sals w  adoption of a f ilter  temperature specification1  of  248 ±  25*F.
Sj *n9er  c ere ala° made to include  the evaporative  residue of the  Method 5
C?8' cona'!*heS' coramonlv called  the  back  half,  as  part  of  the  total  emis-
IQ  ^aif ^dering this material  to be composed  of condensable matter.   The
^ S°l Of ^ tne Method 5 catch on controlled  sources  is  typically  about  20
bH6nt to    coit»bined front-half and  back-half  totals with a  range of a  few
  t  SJ-£ amore than 90%. 2   Strong objections  were raised  to including  the
  i   *in ? Part °f the particulate emission measurement  and  it was dropped
       f0  ver«ion of Method 5.  However, the back-half  catch has  been
        iC 80rae sPecific industri
        -Local control  districts.
    «nd i  80rae sPecific industries and for general application by some
        t
        c° Measuring Condensable Materials
     etigj          between the vapor and particulate phases of condens-
\jS*  Howe   i8 temPerature dependent and thus will exhibit seasonal varia-
Ss toc »e V6C'  from a regulatory point of view, the reference method (s)
    nt 0 8urin9 condensable matter should provide results that are not
    ion j  Ocal temperatures.   Thus some standard temperature for sample
     r th8  needed» such as 0,  5, 20, or 25*C.  Whatever temperature is
    ^ hav6  operati°nal definition for total primary emissions will
     ** Q£6dt0-b* that material collected by the selected method (s) .  This
         "Definition is now used for particulate matter collected by
            °f  Condensable  Matter  Apart from Pre-existing Particles

        a 6nt of  the condensable  portion of  source emissions as a separate
   s    Vfe nt quantity  from pre-existing particles in the stack might be
  •  f'  8°  WayS*  The  first  is  to revive tne inclusion of the Method 5
          C0nd«  a  temperature-controlled coil-type condenser could be
                                  511
-------
                                                                       Llf
     If the back-half catch as described in Method 5 is selected for ^^f
development as a vehicle for collecting condensable materials, certainP
lem areas must be addressed.  The potential for reactions with gases w|  j
could cause positive measurement biases must be eliminated or miniraiz6 ' (
protocol must be devised for carrying out the evaporative process to ^
the large volumes of water from the impinger solutions without incurt*11*^
unacceptable losses of the condensable components that are to be measul to
The potential for reactions of dissolved gases in the impinger soluti° ^,
form artifact particulate material was one argument that was pursued * V
ously during the original debates regarding the inclusion or exclusio"
the Method 5 back half.  Many of the arguments against including the
half concerned reactions with S02 that form sulfates in the i
tions; however, a nitrogen purge of the impinger solutions im
sampling has been shown to effectively prevent this problem.3
     The impingers used in conjunction with Method 5 result in vigor°u
contact between the sample gases and the impinger solutions which m*? g
unnecessarily enhance uptake of such gases when the method is appli®^ tt)i(
sampling for condensable materials.  Thus it may be desirable to li* i  $t
uptake by using a coil-type condenser.  To ensure that no condensate   j$
form of particles escapes with the spent sample gas, a filter is teC° ( i*
downstream of the impinger/condenser devices.  The use of such a ^ .  )!$
called for in the Oregon State Method 7 for condensable matter1* and l
Method 8.2
                                                                      x
     Integration of the irapinger/condenser methods for measuring con
matter with the two current PMj 0 methods under development is reason^
straightforward.  In the case of the constant flow rate (CFR)
addition of the PMj0 cyclone to the Method 5 system previously
all that would be needed for hardware modifications.  Alternatively'
filter could be moved from the Method 5 oven to immediately folio" fc  ^
cyclone for operation in a Method 17 configuration.  For application
Exhaust Gas Recycle (EGR) method,5 few changes would have to be mad*'

  Measurement of Condensable Matter Together with Pre-existing

     To assess the condensables portion of the primary emissions
process correctly with respect to particle size, the actual condens
process that takes place as the emissions mix with ambient air fflust'{iii9 .j|
                                                                       *
replicated.6  A method cannot be devised that would duplicate the
process of flue gas with ambient air for all of the conditions tha t  ^
occur.  However, methods that can reproduce the important features  ^ys
mixing process for some typical conditions have been devised.  All   (
the dilution of a sample gas stream that has been maintained at of ^   ,
stack conditions with ambient air that has been conditioned to va*^t0 f*j,
extents*  In some cases, the dilution air has simply been filtered  Q\\*
existing particles*  In others its temperature and humidity are c°fl Qi
and in some cases scrubbing of vapor phase components is done.  ***
these systems have been reviewed by Pan.7
     Homogeneous nucleation and condensation are unlikely to be * *^
because flue gases and ambient air have sufficiently high concentr  ,0(i  .
small particles to act as primary sites for condensation.  Con^etlace.6
tends to take place at a constant mass rate per unit area of su  *nt **
aerosol number and surface area distributions for almost all afflb * ^
sols, and especially, of emissions from stationary sources, are *
concentrated in the small end of the aerosol size spectrum.  ThaS
onset of condensation, the bulk of the condensation will take

                                   512
-------
    *"* smaller particles of the original aerosol.  The growth rate of the
     Particles is great in terms of change in surface area per unit mass
   .  ted; thus the overall mass transfer from vapor phase to particulate
LOC «jCan be expected to be dominated by particles that were initially in
COQ,]. "^"Particle tail of the distribution.  Thus, for practical terms, the
Inj^j Cables can be considered to be associated wholly with what would have
Of tx* y been tne fine-particle portion of the original size distribution
     Particulate emissions.  This association of condensed matter with the
            end of the aerosol size spectrum is fortunate for the task at
           >f difficulties in transporting aerosols without incurring
          losses of particles in the larger size classes.  Because the
    ^ jnust be transported to a location outside the duct for a dilution
       system, the initial size distribution of the particles in the flue
         be completely duplicated in a system that mimics the mixing of
        with ambient air.  On the other hand, systems can be designed that
          low enough in the critical size range (a few hundredths of a
      et to a few micrometers) to enable the use of a method involving
          an extracted stack gas sample.

 * Plim[era* dilution systems have been designed and constructed to simulate
 ...  "»« dilution process for collecting aerosol samples that contain
          material (s)  in the particulate phase.  Most of these systems
         ped to investigate health effects of emissions or to obtain
  .  »   e source emissions data (fingerprints)  for use in receptor raodel-
°t the -C0raraon Problem in a11 of these systems is that the relative humidity
C*u8e<3 K     diluted gas mixture must be below saturation to avoid problems
     b¥ wetting the filters.
fill      systems. only those of Boubel and Ripper ton,9 Heinsohn et al.6
K  9r» •   80n et al«    collected the entire diluted sample by filtration
1 5h-voilnietric analysis'   Tne Boubel and Ripper ton sampler was based on a
i ' the     stack sampler  and used relatively modest dilution.  Consequent-
  e       l sample temperature was high.   The Heinsohn et al. system was
    t  £°r quantifying total primary emissions from stationary sources.
    tf8ults obtained with the system show significantly higher emission
       n those measured with conventional methods.  Development of the
   t c*tts discontinued because sample gas flow control became a problem
:» vas6rtaln sampling conditions.  The system described by Williamson et
 °c   U8ed to measure the condensable matter in emissions from a number of
I ftt ^ durin9 the development of the Inhalable Particulate10 and the
t*ll*9l«l° emissi°ns factor  data bases.  However, the size of the system and
 0( 9«rt  P°int measurement protocol used in applying it made it undesirable
.    "ral use.
          Methodology
   is
u'^end lp ensure  that the most appropriate and acceptable method was
th^*t«l ?ed» a  number  of people in regulatory agencies at the state and
  th$ «jevel'  experts in source measurement and characterization,  experts
      **ld of pollutant transport and fate, and experts in the field of
   ->;h *"lplin9 and analysis were contacted for advice and ideas.   The
°?*n Condpte£erred by  a majority was that of air dilution cooling rather
i(hii%Mlenaation in impingers or condensers.  From the standpoint  of ease
           it was  generally agreed that the Method 5 back-half approach
»wcon(J  pr«£erable.   However,  the difficulties posed by the large volumes
St *ion Jed water  and possible reactions of dissolved gases in aqueous
   r    ° f°rm artifacts that would erroneously be counted as condensable
         most respondents to prefer the air dilution approach.   Also,  in

                                513
-------
 the dilution approach, the sample is collected entirely on a filter.  jJJ
 source samples taken with a dilution system would be expected to provl^
 closer match to ambient samples taken downwind in the source plume, *,tjl
 the data more directly applicable to assessing source impact.  Hence/ '.
 recommended that the air dilution approach be selected for developm®11*
 the EPA method for measuring primary source emissions.

                   Working Definition of Condensable Matter

      Because the split of any condensable component of emissions tt'&fj
 source between the vapor and condensed phases depends on temperature 'J
 for some compounds, on the amount of dilution that has taken place *n ^
 ground concentrations of the material), it is recommended that total rj
 emissions be defined as the material measured by the sampling train 8* J\
 ,by EPA for collecting condensable matter. • Just as the particulate «•*.{
 for regulatory purposes have historically been defined as the front-*1* $
 catch of Method 5,  primary emissions that include condensable matter *\
 be defined as the catch of the dilution sampling train,  inclusive ot *
 nozzle through the filter.

                           Suggested  Implementation

     A sketch of  the suggested sampling system,  identifying  key  Paft8i
 shown  in  Figure 1.   To  minimize the  size and  power requirements  of  *%)
 system for  use  in the field,  a total flow of  about 280 liter/min (10 !/
 suggested with a  sample flow of about 14 liter/min (0.5  cfm).  This *
 flow is comparable  to but  somewhat less than  the  flow normally used *
 Method 5  sampling  (0.75 cfm).   The cyclone shown  at the  probe inlet *
 provided  to hold subsequent  deposition  of particulate matter  around *
 sample flow meter to acceptably low  levels to ensure that it  functi^ ^
 properly.   A cyclone which removes particles  larger  than nominally ^'^,
 aerodynamic diameter is recommended  for this  purpose on  the basis °f ^
 ence with similar systems.   This also provides a  convenient cut  f°* $$
 ing total primary emissions  of fine  particles if  they are defined »* $
 nominally 2.5 gm and smaller.   By mounting a  PM10  cyclone  upstream °^/
 cyclone illustrated, a  measure of PM]0  emissions  that includes the G $
 able fraction can be obtained.   Recommended conditioning for  the di1 to'
 air is filtration to remove  background  particulate matter, and dryi"^ J
 dewpoint of  35*F  (1.5*C),  with  a final  temperature of 68*F  (20*O-  &f
 dewpoint  and moderately high dilution ratio should prevent problem9
 moisture condensation for most  flue gases  that might be  encountered'^^
 Because the  sample flow is only  a small part of the total flow* ne** j,«'
 of  the sample flow at the exhaust end of the system, as  in Method '/ ^
 possible  in  the recommended setup.   Instead, a flow-metering eie^e0^
 installed immediately upstream of the sample inlet to the dilutee-  .
 venturi-type flow meter is recommended, but other types might be "8
 Method 5 dry gas meter cannot be used to obtain the integrated s
 volume, but  an electronic  flow totalizer using the signal from the
 flow meter could be  used for that purpose.

                               Sampling Protocol
     The suggested sampling protocol for a measurement of total
including condensable matter, parallels that of Method 5 in most
A nozzle of appropriate size for the design sampling rate of the  * t.  .
the gas velocities in the stack would be mounted on the cyclone *n   v
standard Method 5-type traverse pattern would be used in
sampling rate would be set according to the venturi meter

                                   514
-------
fcjj
s*nipli   fls needed to maintain isokinetic  sampling conditions.   During
»oiu_ n9 an electtonic  integrator would be used to measure  the  total gas
  j   SamPled.  Although  this protocol does  not maintain a fixed dilution
   I4? 6 variability  induced should not  pose a problem*   A minimum accept-
    . lution factor should  be specified  (which would in turn set a maximum
       rate) ; the nozzle size to be used for a traverse would then be
 oi^t   to ensure that the  minimum dilution  specification  would not be
 VCI.

. °ktain PMJO  emissions data  including condensable matter, a  PM
  .Q.
                                                                      10
  catt C*n be added to  tne train.  The CFR protocol5 could  then be applied
  inj cy Out tne traverse  at the fixed sampling rate required to hold the
                                                                    PUMP
                                                    EXHAUST l>10clml
                                                          ORIFICE
                                                          METER FOR
                                                          TOTAL FLOW
                                     VENTURI
                                     FLOWMETfR
                                                            fLEXIBLE
                                                            HOSE
                  PROI
                                                      DILUTfR
                                                       MIXING
                                                       ZONE
                           PITOT
                           MANOMETER
                                   SAMPLING
                                   RATE
                                   MANOMETER
                STACK
                WALL
 FILTER

OHIFICE METER
FOR DILUTION
AIR FLOW
                                                       FLEXIBLE HOSE
                                                                         DILUTION
                                                                         AIR IN
    '9Ure 1.  Suggested sampling system for the measurement of primary paniculate
           emissions, including condensables.
 W.*h* d.
 V &ti»     on metnod appears to be the  approach of choice for measuring
 V0*'•    Y emissions including condensable matter from  stationary
    tot ,7°natruction of  a system with traversing capability that is suit-
    *°  £leld use appears feasible; consequently, this  is the recommended
        °c development as a reference method.
                                    515
-------
       The most promising  alternative  is  the  irapinger/condenser
 The  latter requires  the  least  investment  in terms of methods
 capital equipment  costs, and operator training.  On a technical ba»i|(
 however/ this approach may not be as sound  as that of air dilution*

 Disclaimer

       This paper has  been reviewed in accordance with the U.S.
 Protection Agency's  peer review and administrative review policies **p
 approved for presentation and  publication.  The information contain*",
 this  paper does not  necessarily reflect Environmental Protection A9*11
 policy.

 References

 1.    U.S.  Environmental Protection Agency,  Office of Air and Waste
      Management; Office of Air Quality Planning and Standards, m
      Comment Summary:  Revisions to Methods 1-8 in Appendix A of
      of Performance for New Stationary Sources," U.S. Environment**
      Protection Agency, Research Triangle Park, NC (June 1977).
9.
 2.   Emission Standards and Engineering Division, "Estimation of &*#
      Importance of Condensed Particulate Matter to Ambient Particul*
      Levels," BPA-450/3-81-005a, U.S. Environmental Protection Ageiwj,
      Office of Air Quality Planning and Standards, Research Tri«n9**
      NC (1981).

 3.   D. R.  Kendall, "Recommendations on a preferred procedure tot **'f
      determination of particulate in gaseous emissions," J. Air  f^f^
      Control Assoc.,  26:  871  (1976).

 4.   Oregon DEQ,  "Source  Sampling Method 7 for  Condensable Emia«»ion"
      Department of Environmental Quality, Salem,  OR (Aug. 1981) •

 5.   Federal Register  53, Ho.  68, Proposed Rules  (1988).

 6.    R.  J.  Heinsohn, J. W.  Davis, K.  T.  Knapp,  "Dilution source  *
      system,"  Environ.  Sci* Technol.;  14:1205 (1980).
7.   Y. S. Pan, "Review Summary of  Stack  Sampling of  Pine  PafticU^
     DOE/PETC/TR-87/2, Pittsburgh Energy  Technology Center,  Pitt*b
      (1986).
8.   Dale R. Warren, J. H. Seinfeld, "Simulation of aerosol si*e
     distribution evolution in systems with simultaneous nucle»fci^jj)'
     condensation, and coagulation," Aerosol Sci. Technol. 4** '
R. W. Boubel, L. A. Ripper ton, "Benzo(a)pyrene production djj* (\r
controlled combustion," J. Air Pollut. Control Assoc., t3i55
-------
        MATTER-ORGANIC COMPOUND INTERACTIONS,  MUNICIPAL INCINERATOR FLY ASH
          Jr-.  D-  E-  Wagoner,
               Li  s     f
  ac
      Triangle  Park,  NC 27709
  • M
   tyaf8eson.  J.  E.  Knoll,  M.  R.  Midgett
   oi^a  urance  Division/Source Branch
K tfid Stnta^"  ^onitoring and  Support Laboratory
           Environmental Protection Agency
               Park,  NC 27711
   Pi
  ^a ma?Sll)  a  fine  particulate effluent from municipal Incinerators,
    5% ^ by"Product  of this combustion process.  Fly ash consists
        n°rganic material.   The large surface area of the fine fly
         t:  is a site  for adsorption and concentration of
         ""utagenic and carcinogenic polycyclic organic material

   ft
   ate i  nS  techniques used to collect organic material from par-
  s oj; q*detl combustion gas involve hot (125°C) sampling through a
   Ol: qu*rtZ  probe followed by removal of the particles on a glass
   at thiCt2  filter-   Material (organic or inorganic) that is gas
    Ol^ XAn  5^>6Vated te°Perature can pass through the filter and
              sorbent resin.   However, this gas-phase semivolatile
 K
   ^d 0   Soinetimes  be  lost through interaction with particles
       n t«  filter.

       f Cal"  nethods commonly used to study recovery of organic
   *n appr°m particulate matter require extraction of the compounds
         0priate  solvent before analysis.   The most widely used
         ^6  been Soxhlet or ultrasonic extraction.
   s y of tJPtion  of  POM on particles can be irreversible if the
   °*Ul)tlit   lndlvidual POM for a particle is much stronger than
          y °f  the POM  in the extraction solvent.
                                   517
-------
     The development of a laboratory based gas chromatographic
approach for evaluating the effects of compound/particle matter
interaction using temperature and flow conditions which model partide
and organic material collection in the Semi-VOST method is described.
Experimental results exploring the gas phase retention of selected
organic compounds on the Municipal Waste Incineration fly ash are
presented.

  In summary:

     o    Organic compound interaction with two discrete samples of
          municipal waste incinerator fly ash was different, fly ash
          sample 2 appearing slightly less active than ash sample 1-

     o    Organic compounds interacted with unsilanized glass fiber
          filter material, which retained the chlorinated phenol
          compounds significantly.

     o    Interaction occurred between fly ash and several compounds
          in the test set.  These compounds include
          2,3-dichlorophenol, 2,3,4-trichlorophenol, dibenzofuran,
          benzo[e]pyrene, and 2,2r,6,6'-tetrachlorobiphenyl.

Introduction

     In the first phase of field tests of the Semivolatile Organic
Sampling Train (Semi-VOST) method, selected compounds were spiked
dynamically after the particulate filter and before the XAD-2  res!*1
to preclude interactions between the compound and particulate matter
on the filter.  Laboratory tests were then performed to evaluate the
vapor-phase interaction of selected organic compounds and particulac
matter.  Compounds evaluated include toluene, pyridine,
dichlorobenzene,  o-xylene, trichloroethene, naphthalene,
benzo[e]pyrene, 1,2-dichlorobenzene, 1,2,4-trichlorobenzene,
2,3-dichlorophenol, 2,3,4 trichlorophenol, benzofuran, dibenzofuraOi
and perchloroethylene.  These compounds were chosen because they     y
represent a wide range of vapor pressure and polarity and because &  ,
are precursors to more toxic or mutagenic compounds found in       "
waste incinerators.

Experimental

     A Shamatzu 6-AM gas chromatograph (GC) was modified to all°w
liquid injection of high molecular weight compounds with low
pressures either directly onto the fly ash particles or through
bypass directly to the GC column (Figure 1).  Cartridges to hold
particles and bypass valve were housed in a temperature-controHe
-------
               ?n theT   I *" T "° that the C°mp°Und °f
               J?     ?   P?aSe and S° that the Section solvent was
                       yte   f0re 1C 6ntered the GC flam
                                         set so that the
 ,  "citv t-Kv   u  i_     —   — -• = «= was set so cnac tne linear
 J the Se2!°± '^ Particulafe Cartridge was the same as that found
 Jhe gas c^       yS^em Derating with a 100-mm filter.  The use of

 f?^ifai^
 sk   ulatp m^t-f-     -n-j        ween vapor-phase organic compounds and
      " -»wJ                J *  A LIC lyWU
      fiber fn?ai Wafe.lnci^rator fly ash samples, unsilanized
      a*e   f  te  material- a"d quartz wool were all evaluated f
         ne  f or                            samples were evaluated
       to L     ganl° C°mP°unds behaved differently when they were
                                            The incinerator fly ash
                 fiber f M                   ed f°r theSG t6StS   The
                 i          er materlal and q^rtz wool were tested to
St.   "'ine if *.>,   i       	-""wv-i.idj. aim quartz wool were
b^*V°ST mLu ! 8;aSS flber collection substrate used in the
        method affected compound recovery and if the recovery could
        Q by changing to a more inert material like quartz.

                -es of the test compounds were injected through the
                particulate oven to establish particulate-free
         A sequential injection of the same  quantity of test
         s then injected through the particulate-containing path of
                Pparatus.   The  nercent rp^nvofir nf ira^n?- «u	.	^
    ery
    ^d'
 lnte,_nd was used to Mt.aK1,BK fche extent"rcoSo^nd'particulIte
             cussion
    On
    xuajj^j.
   'E f^^     ye data for most of the compounds were acquired from
           ation report, and chromatographic data were integrated bv
 it  ~*y inte     rau°r'  Chromatogranis of several compounds were
coh  (cQlleot-B^ate   y usinS reconstructed digital time and intensity
   Unds In  i   3S PeSk Profiles),  as shown in Figures 2-4.  These
         nciuded the phenols and benzofurans.
8J)  ^coye •
O ltl Tab!"eST°f teSt comP°unds  for each substrate material are
ifc/°tfc«^ foie  I-   These  manual integration calculations  were
         re W^11"shaPed compound  profiles and  compared  to automated
           suits  to  validate the  manual  integration approach
         utomated integration agreed completely  for well-shaped
         manual  results  were  used if profiles  could not be
         automatically.
                                519
-------
      Also  shown in Table  I  are  the  results  of  compound interaction
with  Reeve Angel 934  glass  fiber  filter media  and with quartz  wool.
The weakly basic nature of  the  glass  fiber  filter media is  shown by
the loss of both chlorophenol compounds.  These  results indicate that
sampling for  chlorinated  phenols  with glass fiber filter media will
result  in  lower recovery  compared to  quartz or silanized glass fiber-

      A  second set of  laboratory experiments was  performed to improve
the recovery  of compounds showing losses  after passing through
particulate material  at Semi-VOST conditions.  The  second set  of
experiments was performed at a  more elevated temperature (162  C) in
the particulate oven  box.   Results  of these experiments are given i11
Table II.  Recovery improved dramatically for  dichlorophenol ,
2,3-benzofuran,  and dibenzofuran.   Trichlorophenol  and benzo[e]pyrefle
were  not detected at  the  higher oven  temperature.   This experiment
shows that improvement in recovery  of higher molecular weight
compounds  is  possible at  elevated Semi-VOST oven temperatures.

Conclusions

      The laboratory gas -phase experiments were intended to  evaluate
the interaction between organic analytes  and municipal waste
incineration  fly ash.  Several  compounds  of interest which  are
precursors to hazardous products  of incomplete combustion,  have  been
shown to interact with and  become lost to particulate  matter.  The
laboratory data generated in this study were used to select five
potential  problem compounds and three control  compounds for a  field
test  where problem and control  compounds  will  be dynamically spifcg(*
into  a  dual -probe Semi-VOST system  operating at  a municipal waste
incineration  site.  Problem compounds were  selected because they we
not fully  recovered in the  laboratory tests.   Control  compound
compounds  were selected because they  were fully  recovered in the
laboratory gas -phase  tests  and  because they generated  Gaussian pefllc
profiles,  which implied no  particulate interaction.  The compounds
chosen  for the field  test are shown in Table III.

Disclaimer

                                                                 i
     This paper has been reviewed in accordance with the U.S.
mental Protection Agency's peer review and administrative review
policies and approved for presentation and publication.
                                   520
-------
                                  TABLE  I
                PERCENT RECOVERY  UNDER SOU-TOST COHDITIONS
Fl;
Toluene
Pyrldine
D Ichl or ob enzene
o-Xylene/TCE
Naphthalene
1 ,2-Dlchlorobenzene
1, 2, 3-Trlchloro benzene
2,3-Dlchlorophenol
2 , 3 , 4-Triehlorophenol
2,3-Benzofuran
OLbencofuran.
Bento ( e Ipyrene
2,2' ,6,6'-Tetrachlarobiphenyl
7 Ash
»1
81
103
93
92
95
103
90
20
a
53
0
0
62
Fly Ash I
#2
HA
HA
HA
HA
89
103
89
73
0
92
86
NA
DA
ilruilanized
Glass Wool
NA
HA
HA
HA
HA
HA
HA
Q
0
99
94
HA
NA
Quartz Compound
Wool
NA
NA
NA
NA
HA
HA
HA
35
111
81
115
HA
NA
 HA - Hot analyzed
                                  TABLE II
                    PARIICULATE-ORCAHIC IHTERACTIOH IN
                           HIGH-TEMPERATURE QVEH
Coopcuod
2,3-Dlchlorophenol
2,3,4-Trichlorophenol
2 , 3-Benzof ur an
DLbenzofuran
Benzo[e]pyrene
Percent
123°
20
Q
53
0
0
Recovery
3*0"F
130*
0*
102
117**
0*
* Broad peak resulted  In unreLiable  Integration.
** Initial  Injection. 501 lover than  subsequent injections
                                 TABLE  III
                         COMPOUNDS  SELECTED  FOR FIELD
                    DYNAMIC  SPIKING -  SEMI-TOST VALIDATION
       Not Adsorbed                                     Adsorbed
   naphthalene                                       2,3-Dlchlorophenol
   1,2-Olchlorobenzene                               2,3,4-trlchlorophenol
   1,2,3-Trtchlorobenzene                            Dioeiuofuran
   2,3-Benzofuran                                    Beo*o(ejpyrene
   2,2',6,6'-Tetrachlorobiphenyl
                                    521
-------
               INJECTION PORT
          PARTICIPATE
          MATTER
VALVE
            170
I— FID
       GC OVEN
                      Figure 1  Schematic of  GC
                      particulate  exposure apparatus
            2.8
                      5   7  9  11  13  15  17 19 21  23  25 27  29  31  33
                                       seconds
                      Figure 2   Methane peak  profile
                       through bypass  and  particulate matter.



»s
Respons
rhouson
^^


^

2
2
2
2
T
2
j i -
2.5 -

49 -
48 -
47 -
46 -
/-\-N
y \
/ \ •••
/ v\
r V
J area = 8818.77 *\
I \
«- . / \
« y "w\/ u
-i *i A
^>OO 5SQ 6OO 6.5O 7OO 75O
-------
VALVE
       GC OVEN
           2.2
                                      7  ffi^J
                     Figure 1   Schemotic  of GC
                     porticulote exposure  opporotus
                                                                                          13  15 17  19  21  23 25  27 '29 '3'!
                                                                                             seconds
                                                                            Figure 2   Methane peak profile
                                                                             through bypass  and particulate  matter.
             720
                                            920
                                                            1020
                       seconds
Figure 3.  Dichiorophenol  peak profile through bypass.
                                                                                  500
                                                                                             550
                                                                                                                           700
                    600        650
                       seconds
Figure  4  Dichiorophenol peak profile through flyash
                                                                                                                                      750
-------
MEASUREMENT OF ETHYLENE OXIDE EMISSIONS FROM HOSPITAL STERILIZERS
P. T. Leclair, J. L. Steger, and R. F. Jongleux,
Radian Corporation
Research Triangle Park
North Carolina

W. R, Oliver
Radian Corporation
Sacramento, California

D. A. Levaggi
Bay Area Air Quality Management District
San Francisco, California


     Source tests were conducted at several hospitals with centralized
ethylene oxide (EtO) sterilization operations.  The objective of the tests
was to quantify the EtO emitted to both the air and water during the
facility's normal EtO sterilization and aeration cycles.

     The sampling and analytical methods employed during the hospital
sterilizer source tests were traditional techniques modified to fit the
constraints of the hospital sterilizers tested.  The primary air sampling
methods used were gas velocity measurement and integrated canister
sampling.  The primary water sampling methods were grab sampling with
void-of-air (VGA) vials and gravimetric determination of the water
discharge rate.

     All air and water samples were analyzed on a gas chromatograph with a
flame ionization detector  (GC/FID).  Gaseous samples were introduced into
the GC/FID through a heated gas sampling valve and water samples were
injected using a syringe.

     Average measured air  emissions of EtO ranged from 15 to 34% of the
total original EtO charge  and average EtO emissions to the water ranged
from 22 to 55% of the EtO  charge.  In addition, approximately 20% of the
initial EtO charge was unaccounted for.  A portion of this amount was
expected to be absorbed as residual EtO on the sterile product.  Additional
sources of these nonspecified emissions include fugitive emissions, poor
capture efficiency in the  core room ventilation system, and sample
collection and measurement bias.
                                   524
-------
     Descriptions

sternac^ai? conducted source tests at three hospitals with centralized EtO
the J12a^ion operations with the objective to quantify the EtO emitted to
5erat/ and water during the facility's normal EtO sterilization and
   Ljon cycle.   The sterilizers tested were:
    •   Hospital  A:  a 7-yr-old Castle-Sybrun with a hooding system;
   3    Hospital  B:  15-yr-old AMSCO with an Envirogard system;
   ,'   Hospital  C, no. 3:  a 1-yr-old AMSCO with an Envirogard; and
Uhere D   H°spital  C, no. 4: a 15-yr-old AMSCO with no collection system.
^lUst     ^6j  air samPles were taken from the sterilizer and aerator
    ts and water samples were taken prior to the sewer drain.

SN Methods
  *rann 9as  velocity  was measured with a miniature s-type pitot tube and a
Varie ane  lnclined-oil  manometer.   Open air exhausts were measured with a
      "lometer,  where possible,  for confirmation.

%1 Ca!:?rated  air  samples  were  collected in Summa® treated stainless
pi.  i^1stei"s  by pressurizing evacuated canisters using a metal bellows
  "^dlc  SamPling  rate was controlled using either a restricted orifice
     ie valve  assembly.
        9 the  evacuation  cycle,  water samples from the vacuum pump liquid
        Astern were  collected with minimal  headspace in VOA vials and
      d    at  4°c  in tne  laboratory until  analysis.   When possible, we
    ed tna5Ples at re9ular intervals.  Aliquots of each sample were
   *9e rn?   rm a composite sample that was  analyzed  to determine an
      cotlcentration.
pU(nJc'uid
  / Was measured gravimetrically.  The  plumbing  was  modified at
   a
              flow rate  discharged from the single pass water-sealed
                                      .
    to   and  c  to  temporarily divert the pump discharge from the sewer
         ~° 'lection  bucket.   We weighed  the bucket on a scale over a
  jt  9*1n  peri°d.   The  discharge rate  was calculated from the measured
    eme " and the  elapsed time.  A series of liquid flow rate
    s*mplp Was  made  daily during the sterilizer evacuation sequence.
           Were collected in between flow rate measurements.

    1tJl Methods
    All aiy%
 Wd VUK   d water  samples  were analyzed on a Varian Model 3400 GC/FID
 \VJ°on rS  ? 6"ft  °y 1/8-in.  stainless steel column packed with 1%
     ^'n.  J;J°Pack  B  (60/80).   The nitrogen carrier gas flow rate was
             analyses were performed isothermally at 45 C.
         cai?kples were  injected via a 1  ml  gas sampling loop heated to
      d eth      ed  tne  GC/FI° daily from °*5 PPmv to 50° PPmv Usin9
                oxide gas  standards certified to plus or minus 2%.  All
                                 525
-------
standards and samples were  injected until two consecutive  peak areas were
within 10 percent.  Samples higher than 500 ppmv were  held and re-analyzed
using a calibration curve which ranged from 500 ppmv to  50,000 ppmv  (5%).
Several canister samples, collected under vacuum due to  short  sampling
episodes or high expected concentrations, were pressurized with Grade 5.0
nitrogen prior to analysis.

     A quality control  (QC) canister sample of 390  ppmv  EtO was prepared
using the 5% EtO standard and grade 5.0 nitrogen.   This  QC sample  was
analyzed on four separate days.  The absolute difference between the
measured and predicted  theoretical concentrations ranged from  +1 to +9%
when the system was calibrated from 0.5 to 500 ppmv and  was -14% when the
system was calibrated from  500 to 50,000 ppmv.

     Water samples were introduced into the GC/FID  using a 5-  1  Hamilton
syringe.  Calibration standards were prepared by injecting EtO into a vial
containing 1 ml of deionized water.  Standards and  samples were injected
until three consecutive peak areas were obtained that  were within  10%.  The
GC/FID was calibrated from  90 ppm to 18,000 ppm.  Samples  higher than
18,000 ppm were diluted and re-analyzed.

     A QC sample of 538.1 ppm was prepared using the standard  preparation
procedure.  Analysis of the QC sample resulted in an absolute  difference
between the measured and actual concentrations of -8%.

Calculations

     The total amount of EtO charged to the chamber was  based  on 12% of the
total weight of sterilant gas used.  The sterilant  gas,  commonly called
12/88, was a mixture consisting of nominal concentrations  by weight of 128
EtO and 88% dichlorodifluoromethane.  We calculated the  EtO emitted to the
air from the total volumetric flow passing the sampling  point  over the
sampling interval and the integrated sample concentration.

     The volumetric flow rate in dry standard cubic feet per minute  (dscfm)
was determined from pitot tube measurements using calculations based on EPA
Reference Method 2.  We estimated the moisture content of  the  vent stream
from the vapor pressure of water at the flue temperature,  the  percent
relative humidity of the flue gas, and the pressure in the flue.   We
assumed the relative humidity of the vent gases was 50%  and the vent
pressure was atmospheric (760 mm Hg).  Since the sterilant gas contributed
less than 2% to the vent gas stream, the vent gas molecular weight was
approximately same as air (29 g/mol).

     The EtO concentration  in the air samples was determined by comparing
the response of the samples to a linear least squares  calibration  curve
obtained by analyzing four  standards.  From the average  peak areas obtained
from the analysis of each standard and the corresponding standard  concen-
trations a linear least squares fit of the data points was obtained.  The
average peak area obtained  from the analysis of each sample and the linear
                                   526
-------
     uares slope and y-intercept were used to calculate the sample
             ^or samP^es that were pressurized before analysis, a
    n factor was calculated by dividing the final pressure in the
  j0   °y the initial pressure in the canister.  The original concen-
  mea °f t^1e samPle before dilution was then calculated by multiplying
   eftsured concentration by the dilution factor.
          emitted to water was calculated in one of two ways.  For
     taken at equal time intervals:
     -   .           EtO = Water x Composite/1,000,000           (1)
     nf  is tfle wei9nt °f EtO emitted to the water in lb, Water is the
     th wa*er entering the drain in lb, Composite is the concentration of
g/g   £ne water in  g/g, and 1,000,000 is a conversion factor  in
     Or samples taken at irregular intervals:

S,  AVMPD -                Et° =  tm-   EtO concentration in the water as to calculate the  EtO
 S c*lcul  " in tlie a^r samP^es-  TRe EtO concentration in the standards
   the     ^ ^rom tne barometric pressure, the volume of EtO  injected
         6^ ^e m°lecijlar weight of EtO, the gas constant,  room
        >  and the volume of water into which the EtO was injected.
    al A6su^ts are provided in Table I.  For the first two tests  at
     WcWe did not include the air emissions in the conclusions because
      * EtO concentrations measured during the evacuation cycle  which we
   ,„,,,  ulted from sampling or measurement error.  Also not used in  the
 %J1ta1 A  Were the the emissi°ns to the water during the fourth test  at
 V^g test /Itnou9n the estimated sterilant charge increased by 60%
   !ne
-------
                 '                                                  n OP
     Low aqueous emissions from No. 4 at Hospital C may be due to an ^
drain that possibly allowed greater volatilization of the EtO before
water entered the drain.

     The unaccounted for portion of the initial sterilant charge was
probably emitted to the air by fugitive (unmonitored) sources, but
potential for measurement and sampling error cannot be discounted.

Conclusions
                                                                   e&
     The variation observed in the emission pathways appears to h&*
primary causes.  First, we observed a large variation in the venti a
EtO emissions from the  individual hospital sterilizers. Two sterili*
were equipped with the  hooding system manufactured by American Ster
Company; one sterilizer was fitted with a hospital developed venti
system, and the other sterilizer had no venting system.  Also, *"& j.
drain-liquid/gas separator-vent system was different in each hospij
Second, the EtO emission rates at several potential fugitive emisSL|
sources were not monitored.  Due to the variability in sterilizer a
ventilation systems at  the test sites, all potentially important e»
sources were not identified a priori.
                                                                   t|i
     In summary, the hospital sterilizer source tests conducted f°r
study indicate that of  the total EtO used:                         e
     1.   15 to 34% (mean 25%) is directly emitted to the atmospne'
          sterilization and aeration;
     2.   51 to 55% (mean 53%) is emitted to the water effluent
          (downstream from the liquid/gas separator), based on da*
          two sterilizers with vented or enclosed drains; and
     3.   Approximately 20% is unaccounted for in the field i"
Based on a review of the sampling and analysis methods used, we
that probably the additional EtO is emitted as a fugitive source j J
the hospitals' ventilation systems because water emissions occur   ,$>
single point, at which  measurements were taken, but fugitive air flfliwj
could potentially occur from many points that were not directly,"1 ^fj
Also, we expect a portion of the unidentified EtO charge to be in
of product residuals, most of which would eventually be released
hospital.  Thus, the fraction of EtO accounted for by aqueous
be estimated to be approximately 55% for sterilizers that have
enclosed liquid/gas separators and drains.  The amount directly *
the atmosphere is about 25%.  The remainder may be emitted as a T
air emission source through the building's sterilizer ventilation^
general ventilation systems or may have been unaccounted for due
collection and measurement bias.

Acknowledgements
                                                                  n
     The authors acknowledge Richard E. Honrath, who directed thj
project; William Gergen who participated  in the source testing;
hospital employees whose cooperation made the  source testing P°s
                                    528
-------
01
M
CO
rst&t.£- f. f/os/'frAL ret.
HOSPITAL
I.D.
TEST
NO. DURING t
IN AIR
A
A
A
A
B
B
B
C #3
C #3
C #3
C #4
C #4
C #4
1
2
3
4
1
2
3
1
2
3
1
2
3
0.49b+
0.87 +
9.87 +
5.43 +
28.84 +
11.73 +
15.86 +
0.28d+
0.26°+
12.15 +
25.87 +
32.86 +
11.32 +
0.55
1.49
3.09
0.79
4.10
1.78
1.77
0.25
0.02
1.48
5.56
3.01
1.45
EtO EMISSION IN
?T KESVL rs TAT PERCENTAGE
PERCENTAGE OF ORIGINAL CHARKF
EVACUATION DURING AERATJON
DOWN
6.47b
73.15
48.59
30.60
70.57C
96.32C
92.03C
62.02
71.73
31.60

21.40
21.84
DRAIN
>C+ 6.35
+ 11.96
+ 28.92
± 7.83
+ 11.65
+ 15.47
± 14.78
+ 22.81
+ 16.53
± 6.25
__f
+ 6.23
+ 6.73
IN AIR
5.
9.
8.
5.
10.
23.
11.
9!
4
4.
6.
87b+
54 +
78e+-
69 +
36 +
13 +
57 +
30 +
flRe+
55 +
27 +
5.80
1.71
2.14
0.92
3.17
2.90
1.69
1.79
0.18
1.18
1 05
0.45
0.84
UNOUANTITATFn
87
16
32
58
- 9
-31
-19
34
26
46

41
60
.16b+
.44 +
.76e+
.28 +
.77 +
.18 +
.45 +
.97 +
.40 +
.95 +
f
.19 +
.57 +
85.61
2.52
14.33
11.85
1.38
4.27
2.34
12.38
5.97
6.71

5.36
11.90
                                 Values  are estimates ± 95 percent confidence interval.


                                 Value based on estimate of initial sterilizer EtO charge.
                               ° measurelentl! bell6"ed t0 "' l0" dUe t0 Uck °f 1nclus1on of 1n1tia1  «ater "Charge  in


                                Samples believed to be unrepresentative of true emissions.


                                Percentage based on estimated data.


                                Data sufficient to estimate this value were not obtained.
-------
EVALUATION OF SAMPLING METHODS
FOR MEASURING ETHYLENE OXIDE EMISSIONS FROM STERILIZATION CHAMBERS AND
CONTROL UNITS AND DETERMINING CONTROL UNIT EFFICIENCY
J.L. Steger, P.T. LeClair, R. Jongleux, and W. Gergen
Radian Corporation
Research Triangle Park, North Carolina  27709

J.H. Margeson and M.R. Midgett
Environmental Protection Agency
Environmental Monitoring Systems Laboratory
Research Triangle Park, North Carolina  27711

     The U.S. Environmental Protection Agency (EPA) is currently considering
developing regulations to control ethylene oxide (EO) emissions from
commercial sterilization facilities.  A reliable sampling and analysis
method for measuring EO emissions must be established.  The method must be
capable of measuring total EO emissions and determining the efficiency of EO
control devices.

     At commercial sterilization facilities the EO is emitted from the
chamber or control unit intermittently, and the emissions vary in intensity
and EO content.  This paper describes a field evaluation of a
semi-continuous direct sampling method for commercial sterilization
facilities.

     The facility chosen for the test used a mixture of 12/88 (w/w)
EO/dichlorodifluoromethane (CFC-12) as the sterilizing gas.  Ethylene oxide
emissions to the atmosphere were controlled using an aqueous
absorption-hydrolysis system.

     Samples of the exhausted gas were continually removed from sample ports
located before and after the EO control unit and analyzed using an on-line
gas chromatograph equipped with dual gas sampling valves, columns, and flame
ionization detectors.

     The volumetric flow rate from the control device was determined from
differential pressure measurements across two restricted orifice plates
installed in parallel on the control unit stack.

     The method was evaluated for repeatability, precision, and usefulness
is determining the efficiency of an aqueous absorption-hydrolysis control
system.
                                      530
-------
   a  5e^oc' f°r sampling and analyzing ethylene oxide (EO) in the vent
evalu t   m a sterilization chamber and a dilute acid scrubber was field
^s d t   an(* ^e me'thod's usefulness for measuring control unit efficiency
Wlut   mined-  Tne u-s- EPA listed EO as a possible hazardous air
to con "*' Creatin9 a need f°r a standardized sampling and analytical method
ProCe!jSl Gently determine control equipment efficiency.  The evaluated
chromat^6 used semi-continuous direct sampling with on-line gas
the    graPnic (GC) analysis.  A throughput efficiency was calculated  using
     "355 fl°w rates measured at the inlet and outlet of the control  unit
         ery that was equated to the control unit efficiency was
  emi     from the weight of EO charged into the chamber and the measured
    ISsions at the outlet of the control unit.

 aCi11tV Description

    i7 f!e^d evaluation was conducted at a commercial medical supply
    aa      facility that has three 28 cubic meter sterilizers that use
          ^e initial E0 Cnar9e to the chamber was calculated by
  ""*   ^erilize^ exhaust is controlled by a Chemrox DEOXX^ system,  a
  d» th I  sc>^ubber, containing a mixture of dilute phosphoric  and  sulfuric
 ^ippejat.hydrolyzes the EO to ethylene glycol.  Each tested  chamber was
     u with an oil-sealed, total-recirculating pump with a gas/liquid
     hi  that exhausts to the DEOXX system while recirculating  the  liquid
     PUmP inlet.
 H6 scrSes We>^e simultaneously acquired before and after the control  unit.
 JP 1l1zeJer inlet sampling location, used to obtain a continuous  sample  of
 jobber V fhamber exhaust, was midway between the sterilizer outlet  and  the
  ubber ]n]et-  The exhaust was transferred from the chamber outlet  to the
       1n'et via a 15-cm diameter polyvinylchloride (PVC) duct.

 $cJkl"leasuUtinuous sample of scrubber exhaust was obtained and volumetric
 >W*er v   ents were made at tne scrubber outlet.  Exhaust exited the
 C level   ical1y through a 15-cm diameter duct that exhausted  1.5  m  above
           T° measure volumetric flow, the stack was modified by
      .    "in- diameter ductwork and two butterfly valves to divert the
   Or>1f1ca  ist tnr«ugh one of the two parallel ducts, a sampling probe,
  iyi      plates in parallel, and wet and dry bulb temperature  probes.

      e°xide Sampling

        ^s were drawn into heated Teflon® lines using Teflon-lined
         p Ps•  Portions of the samples were routed through heated  valves
         I the samples onto the GC columns.  Fine metering valves and
            e used after the valves to control the flow rates of the
                                 531
-------
      Testing began when the DEOXX scrubber started to exhaust prior to ^{
 initial chamber evacuation.  Each test consisted of seven evacuations*
 initial chamber evacuation and pump down and six air in-bleeds and     ^
 subsequent evacuations.  The start time and end time of the evacuation
 identified by flow or lack of flow across the orifice plates.
 Volumetric Flow Rate Measurement

      Nitrogen,  oxygen, water, EO, and CFC-12 were the main components  ^
 sterilizer exhaust gas.  The emissions of EO and CFC-12 were continuou  st
 monitored by GC/FID.  Volumetric flow rate measurements of scrubber &
 were performed  at the scrubber outlet location using two standard,
 squared-edged orifice plates mounted in parallel ducts.  The or^Sue
 diameters used  were 1.763-in. (4.48-cm) and 2.591-in. (6.58-cm).  T" ^ib
 percent moisture of the stack gas was determined by the wet bulb/dry
 method.  Temperatures were measured using a type-K thermocouple and
 pyrometer.   Oxygen emissions were measured with Fyrite oxygen indiea ]sof
 The nitrogen concentration was determined by difference.  Percent ijj,  Of
 oxygen  were usually measured once during each evacuation.  For sever
 the runs oxygen was measured at 1- or 2-min intervals to determine i
 residual  volume of the scrubber system.

 Analytical  Procedures                                                  (
                                                                   rCT-J  n
      The analytical  method used for the measurement of the EO and   i fl"
 gas chromatography with flame ionization detection (GC/FID).   The °g,poP
 Varian  3400 GC  was equipped with a heated valve box containing tv/?o0p ^ft
 valves.   An 0.25 ml  loop was used on the inlet sample line and a ig jO^
 2 ml  was  used on the outlet sample line.  The analytical columns *
 (3  m) x 1/8 in.  (3 mm) O.D. stainless steel  columns containing $%
 on  60/80  Carbopack B.   The FID electrometers were connected to
 »
Shimadzu CR1-A  integrators.

     Both channels  of  the  chromatograph  were calibrated for EO an  afiaiy
the beginning and end  of the  day.   At  least one standard was also
between tests.   Certified  standards were purchased ranging in E centr^
concentration from  less  than  1  ppmv to 20% vol  and in CFC-12 cone
from 1200 ppmv  to 62.5%  vol.
w
Results
                                                                 s c
     The sampling  and  analytical  method  was  evaluated using a.?Ler
containing known concentrations of  EO  and  CFC-12.   The gas cylina
first analyzed on  the  GC.   Then the gas  cylinder was sampled as  ^e tja
slipstream.  Response  of the  cylinder  sample through the s?mP'fiy.
compared to the response of the cylinder sample  analyzed direct  3
                                                                  1 nfl a til
     The inlet sampling and analytical bias  was  measured twice   Q./jjLjJ-
2,508 ppmv EO and  6,022 ppmv  CFC-12 standard yielding ranges JT  f0rV,
average of 3.5% for EO and  4.3 to 12.5%  with an  average of B.«»  usljj jJ
The outlet sampling and analytical  bias  was  measured three tijj  j^nir
502.4 ppmv EO and  1,200 ppmv  CFC-12 standard yielding ranges 01  flf ,*.
with an average of 7.4% for EO and  -9.5  to 4.8%  with an average
CFC-12.
                                    532
-------
  Cted emissions.           ™       emissions were larger than the      '
                        °J ? emitted to the atmosphere from the control
                             ^^^ &S! S !£ 1 ?
              ^
t>  ^Ost  c
Cf 1r> the^ntf-1"0!* *" E0 !mission measurement  probably  resulted from
    6 °* dVeS? n!10" °f the fl°W ^/concentration  prof  e    °m
   *: ^let thin 1? th^cr mKKSUred ^ith  greater P^cision at the
    C                           °!!tiet as was  expected  because of the
                                                       E°
                         USin? ^he throu9hP"t method with the data from
                                                                     1
       ,

     a1alSj-°^fthe throughput and recovery efficiencies were done bv »
     d ftJtiJ  fThVear;na?pC^{fNOVA] witkh sampllng-calculatloM? procedures
 c1ens W3s teitPrf f« 3 Jrac*10n term between the  "Iculational
        -MILS nacnn     *u     -,             a significant effect on
        the   i      on      calculational procedure used   A
     then theapfflar°na\mfthod used has no effect was'
     „ _  'trie erterT u/a« tai/on +« k,» -^	-^.	^
            e  frtu                               .
     I a P of 0 as fn? thtafen*to  b'  Si9nificant-  A one-way ANOVA
                    °      tests "Sin9 chambers which did not contain
                       ons were calculated using the chamber pressure
        cfianqe  in th    calculate ifllet flow rates.   No correction  was
        ^   Stltict? 9?S comP°s1tion that occurred  while  the gas was
        e
-------
significantly different due to the high efficiency of the EO control
Therefore, in tests performed on units that are closed systems; tW^
estimation based on temperature, pressure and assumed molecular wei9n
be a possible alternative to orifice plate installation.
Conclusions
                                                                    i LA
     Five conclusions were based on the field test results.  First,  
-------
                                   J-.   CONTROL UNIT INLET AND OUTLET EO EMISSIONS AND EFFICIENCY
Test
Number

7
9
10
12
14
en
oo 15
en
Average
Stand. Dev.
Rel. St. Dev.
Initial EO
Charged to
Chamber (Ib)

43.8
41.5
42.4
41.5
42.0
41.2
42.07
0.95
2.26%
EO Left
in
Chamber
(103 Ib)
0.42
1.5
0.22
0.16
0.16
0.07
0.42
0.54

EO Entering
Control Unit
Measured (Ib)

24.19a
60.59b
62.12D
44.00
48.80
52.82
48.75
13.87
28.4%
EO Exiting
Control Unit
Measured (Ib)

0.0433
0.011
0.029
0.011
0.021
0.014
0.022
0.013
58.7%
Throughput
Efficiency

99.82%a
99.98%
99.95%
99.98%
99.96%
99.97%
99.94%
0.06%

Recovery
Efficiency

99.90%a .
99.97%
99.93%
99.97%
99.95%
99.97%
99.95%
0.03%


                                     -  •"•" —i" "-a >™>"H um my me rirst lu minutes of the evacuation and the
 FID flame was extinguished during portions of the third and fourth evacuations.  Loss of these samples may explain
 the lower mass of EO entering the control unit during this test.


"The EO standard calibration curve for inlet samples on October 8, 1987 (the day Tests 9 and 10 were performed)  was
 lower than on the other test days.  This would have raised the measured EO concentrations,  and caused  the  EO mass
 flow into the control unit to be over estimated.
-------
FEASIBILITY STUDY ON REAL TIME MEASUREMENT
OF TOXIC INCINERATOR EMISSIONS WITH A TRACE
ATMOSPHERIC GAS ANALYZER
L.E.Slivon, G.M.Sverdrup, W.H.Piispanen, J.E.Orban
Battelle Columbus Division
505 King Avenue
Columbus, Ohio 43201

S.D.Tanner, W.Fisher
SCI EX
55 Glen Cameron Road
Thornhill, Ontario, Canada L3T 1P2


     A feasibility study was conducted to investigate and demonstra
potential use of a Trace Atmospheric Gas Analyzer (TAGA) for momto
incineration stack emissions in a chemical agent demilitarization
operation.  Compounds of interest for detection were the nerve ag
GB and the vesicant HD.   Analytical requirements included reprodu
detection limits below 30 nanograms/cubic meter while maintaining^$i
response time of less than 15 seconds for agent VX.  The feasibil' J
was conducted using Battelle's TAGA tandem mass spectrometer witn
atmospheric pressure chemical ionization (APCI) source.             J

     Simulants were used to model the ionization and collisional *
dissociation behavior of the chemical agents VX, GB and HD.
Diisopropylmethylphosphonate (DIMP) was used as the simulant for  ?
organophosphate agent VX and will serve at the focus of discussio jjj,
the detection requirements for VX were the most stringent of the  ^
agents.  High humidity acidic gas streams were generated in order
approximate the composition and temperature of a demilitarization
combustion stack matrix.  These gas streams were spiked with trac^
of simulant compounds and sampled directly by the TAGA at a rate
liters/minute.

      While the TAGA was set to monitor a structurally character!  .
parent/daughter ion transition for a particular simulant, severa  te$
randomized series of simulant challenge concentrations were ge»;
the gas stream.  The TAGA provided a linear response to simu     ^
concentration and achieved a steady state signal level within! ^s
seconds following a change in simulant concentration.  The TAG£
to clearly distinguish between zero and approximately 0.1 times
allowable stack concentration for each of the simulants tested'
                                   536
-------
                             Introduction
>se of Ihl  S*atesMc?n9ress has directed  the  Department of Defense to
               ire    ted
                    M
                  re  -;-ted !^es st°<*P"e of lethal chemical  agents
           .nitions by September 30, 1994 (1).  The Department  of
         f0^en?ined.that on-site incineration is the most efficien?
      etv of destructlon °f che>nical agents while maintaining the health
  na   ecis?I humai? P°Pu1ations and the environment.   Before making a
       J S °n instrum?ntation for workplace and incineration  stack
      The'off analy+;cal technologies are being evaluated  by the U.S.
       DeSn J6 °f + -e Pr0gram Executive Officer -  Program Manager for
       and d™ an"tion contracted with Battelle Columbus  Division to
       ss Jf ?nStrate th? suitability of a commercially available
         S               ?5e SSiex TAGA'  for real  tl
-------
degrees C (approximately 64 mole percent water vapor), Ng, 02, NO'^to
S02-  Additional agent specific components  (H3P04, HF, HC1) were
the stack matrix.  The matrix was generated at a rate of 90 liter
at a final temperature of 140-150 degrees C.  The matrix was spil;-.  t
simulant and introduced directly into the TAGA APCI  ion source witnou
dilution or cooling.

   The simulants used in this study were Diisopropylmethylphosphorate.^
(DIMP), Dimethylmethylphosphonate (DMMP) and 2-chloroethyl ethyl sun j
(Half Mustard).  These simulants were selected to represent the chem .^
agents VX, GB and HD respectively.  The simulants were individually  L
into the matrix stream at concentrations of 0, 0.11, 0.46, 1.0 and *•
times the ASC of the corresponding chemical agents.  The equivalent  ^
concentrations for 1.0 ASC were 0.03, 0.3 and 3.0 micrograms/cubiC "  .^
for VX, GB and HD respectively  (3).  Accurate and reproducible spi*
achieved using a Sage model 341 syringe drive and a  250 microliter
containing a dilute solution of simulant in hexane.  The hexane sol
was delivered at one of five preset rates (including zero) to a l°.wt
volume platinum capillary thermal vaporizer designed in support of
study.  A diagram of the vaporizer is shown in Figure 2.  Randomly   Of
changes in simulant concentration were rapidly made  by selecting °" ^\t
the syringe drive delivery rates without any other changes to the f»
flow rate, simulant solution concentration or TAGA operational m°  '^
Preliminary studies confirmed previous experience that the hexane
has no effect on the sensitivity of the TAGA APCI toward simulants  /
chemical agents.  The generation of 1.0 ASC of DIMP  (0.03 microgra^
cubic meter), for example, required the delivery of  a 0.131 nanogr*  f$
microliter hexane solution at a rate of 20.6 microliters/minute, v v
into a flowing matrix stream of approximately 90 liters/minute.

                          Results  and  Discussion
   Different sequences of random challenge concentrations were 9..
over a period of several days for each simulant and matrix C9irl')ilnfn W
The limits of detection obtained from these tests are summarized   j to
1.  The limit of detection is defined as that concentration re^ t
                                                                 t\
produce a response that can be distinguished from zero with a c°.noin li"
of 99 percent.  The confidence interval  (+/-3 SD) was obtained tr
regression of the challenge data.   In each case, the TAGA met tn e  , ^
requirement that the limit of detection  be equal to or less th^Mp'ifl a
The actual response of the TAGA to  changing concentrations of DJf
stack matrix is illustrated in Figure 3.  The ordinate of Figure  fftfl $•
represents actual ions per second reaching the detector resulti^^s
structurally specific parent to daughter ion transition for this  tj
-------
                               References

        Law 99-145,  "Department  of Defense Authorization Act, 1986",
       er 8,  1985,  Title XIV,  Part B,  Section  1412.
1 Syerd
Fish  UP'G-M.,  Slivon,L.E., Orban.J.E., Piispanen.W.H.,  Tanner,S.D.,
Phac  |W* and  Urwin.P.T.,  "New Concepts in Chemical  Agent Monitoring-
   6 I - Feasibility Demonstration of the Trace  Atmospheric Gas
   "~~r TAGA  6000E", Battelle Columbus report to the U.S.Army Office
      Program Executive Officer -  Program Manager for Chemical
      arization,  Contract DAAA15-86-C-0106, October 30,  1987.
     ntl
      y revi'sed  ASC values for  VX,  GB and HD are 0,3,  0.3 and 30
  toirr         ise       vaues  or    ,
    ^grams/cubic  meter, 53, FR, 8504.
               Table  1.    Observed Limits  of Detection
Simulant
DIMP
DMMP
ilf Mustard
in Units of ASC
Stack
Matrix
0.024
0.024
0.014
Flue
Matrix
0.082
0.008
0.003
             N, On MembfMW Interface
                                               CAD
                                     (Colllilonally Activated Dissociation)
                                       Argon
                                       Gas                  Detector
                                                         (Pulse Counting)
                                                o*
                                           ("transparent" Quad)
               Focussing    Cryogenic
                        Vacuum Pump
                Figure 1.   Diagram of the TAGA  6000E.
                                  539
-------
VIC" 0010.030" ID
Suinlcsi Steel Tuba
 1'lfSwagelok
Union (modified)
                          /HMtirLiid   xTi(lonS«il
                              StainlcM Steel Shroud
                            Wrapped With Teflon Tape
                                                L
                                                  ReiiiUlK*
    100nm ID
 Platinum Capillary
(from lyringe pump)
                                                      TOTAS*
                    -lOOml/min
                     Clean Air
           Figure  2.    Platinum  capillary  simulant vaporizer.
01
Q.
VI
             Values are ASC concentration
             equivalents for n.  1.0 ASC = 30
                10
20       30       40       59       60

        TAGA Sampling  Interval
                                                                        .  .-
   Figure 3.     Real-time  TAGA response  to  DIMP in  a  stacK m
                                         540
-------
       N THE VOC ANALYTICAL METHOD
  nt USE OF A TOC ANALYZER
,
  C
      terson,  C.  K.  Sokol,  and  R.  K.  M.  Jayanty
         Environmental  Measurements
        r1angle  Institute
       Bangle  Park,  NC   27709
     m    '  E-  Kno11'  and  M-  R-  Mldgett
UA Enw!nta1  Monitoring Systems  Laboratory
 Wh T°nmenta1  Protection Agency
     n 'Bangle  Park,  NC   27711


J?e 1n saHa?"l1clu1d counter-current dynamic  1mp1nger  has  been  adapted  for
s* Sinn  ng Vo1at1le organic  compounds  (VOCs)  from  source  emissions.
Do areite n  method uses an  aqueous solution of  potassium hydroxide  to
l^stHh»r9an1cs  Into  strlppables, which are removed by  the solution,  and
t  1Ppabl  es» Wh1ch  rema1n 1n  tne gas Phase-  The  liquid fraction
c'e gas fles'  ^s analyzed on a  Total Organic Compound (TOC) Analyzer and
c*:toun<| fMMtlon  (non-str1ppables) 1S  analyzed on  a Non-Methane  Organic
th ^tj-ati     Analyzer.  The  results are combined to  give total
  s°Utva  n  °^ VOCs,  measured 1n parts-per-m1ll1on  of carbon (ppmC),  1n
      ge gas.

        P^ger-VOC/TOC system works well  with compounds that  are
        ^etely strlppable (carboxyllc  adds  and alcohols)  or
        J°n-str1ppable (alkanes and aromatic  hydrocarbons).   It works
        wjth compounds that are only partially strlppable (aldehydes  and
     is preliminary  results  Indicate that the dynamic  1mp1nger-VOC/TOC
   nUous      t  as accurate  as Method 25  and  offers  the advantages  of  a
    "9  a^PUng  process, availability of concentration data  during
      1  a"a  excellent  potential  for  automation.
                                541
-------
 Introduction

      The gas-liquid counter-current dynamic  Impinger is  used  to  sepa1"3
 gaseous mixtures into strippable and non-strippable  fractions.   This   y
 sampling system was first reported by EPA in 1982. *   To  date,  its pi"1IB
 application has been to source emission  sampling  for halogenated
      Because of its  potential  for alleviating  some  of  the  shortcoming
 EPA Method  25,  the impinger,  with some  modifications in  design  and
 operation,  has  been  adapted  for use  in  sampling  source emissions of
 volatile  organic compounds  (VOCs).   Obvious  advantages of  the
 impinger  over Method 25  include a continuous sampling  process,
 being replenished with source gas and fresh  stripping  liquid; the
 to  provide  concentration data during sampling; and  the potential for
 automation.

      The  application of  the  dynamic  Impinger sampling  system to the
 analysis  of  total  gaseous non-methane organic  compounds  (TGNMOC) *
 source  emissions involves the use of an aqueous  solution of a
 as  the  stripping liquid.  The basic  solution removes C02 along
 base  and/or  water soluble compounds  from the source gas.   Non-stnPP
 non-methane  organic  compounds (NMOC) in the  depleted gas stream ar
 measured  by  NMOC analysis.   Strippable  organic compounds are measu
 the liquid fraction  by analysis for  total organic compounds (TOC)
 NMOC  and  TOC analysis, all organic compounds are measured  as
 results of the  two analyses  are combined to  give the TGNMOC conc
     The primary purpose of  the  current study was to develop and
necessary methodology  for  collection and  fractionation of TGNMOC ^
using the dynamic  1mp1nger-VOC/TOC  system.  This was to be accomp"   o
evaluating the precision and accuracy of  the system with indivldua ^ t
compounds of several different functional groups and with mixtures ^
compounds and by comparing the performance of the dynamic 1roplpgeLl in
system with Method 25  in the laboratory (with complex mixtures) «"
field tests.

Experimental Methods

                           Description of Method
                                                               f a
     The counter-current dynamic  Impinger (Figure 1) consists OT
condenser with a side-arm  attached  near the main inlet/outlet *l
end.  This provides attachments  for 4 lines:  Liquid 1n (top slae
liquid out (bottom), gas in  (bottom side-arm), and gas out
tube of the condenser is filled with chips of an Inert packing »«  s   ,
held in place by a plug of glass wool.  The inert material decre   , jj
internal volume, slows the ascent of bubbles, and minimizes vert  ^
of stripping liquid within the Impinger.  The water jacket on ^sK*
allows sampling at temperatures other than ambient, if that is °
                                                                   .Cl LI
                                                              ., UA
     Although several plumbing arrangements have been tested, * ^
appropriate one for the current study involves pumping the ^'^e w
bottom of the Impinger at about 1 mL/min and pumping gas frpj * er 1s
the Impinger at about 14 mL/min.  The liquid level in the imp1"^ is"
adjusted by adjusting the height of the source liquid bottle.   eflt JJ,t'
difficult to maintain a constant liquid level with this ar"annei  "•iy
when both liquid and gas are pumped Into the Impinger, but


                                   542
-------
that°f evacuated canisters for collection of nonstrlppable gases requires
  1 sample gas be pulled from the implnger rather than pumped Into 1t.
       mechanism of the separation 1s the partitioning of each component
     mixture between the gas phase and a liquid phase flowing 1n the
     e direction:  Liquid 1n the implnger moves slowly downward 1n the
.     r as tne source gas bubbles upward through it.  Strippable compounds
1n :: with or dissolve 1n the liquid phase; non-strippable compounds remain
  tne gas phase.

an aJn the current application, sampling for TGNMOC  in source emissions,
The st ?US sol"tion of potassium hydroxide is used as the stripping  liquid.
Total I'PPable fraction is collected in glass bottles and analyzed on a
Coiier:r9anic Carbon (TOC) Analyzer.  The non-strippable fraction is
Car^n ti 1n an evacuated Tedlar bag and analyzed on  a Non-Methane Organic
tlie  * VNMOC) Analyzer.  The total concentration of organics  (as ppmC) in
aflalys       source gas is calculated by combining the results of the two


                        Accuracy and Precision

I'*cove55uracy 1n the !aboratory evaluation was determined as  percent
   ss n   Two Parallel Implnger sampling systems were used  1n all runs  to
     Precision.  A recovery of 90% with a percent relative  standard
        URSD) of 10% or less between the two impingers was considered
        •

                      Test Compounds and Sources

        compounds chosen for the study Included three alkanes  (propane,
     ' an
-------
     At the end of the sampling period, sample introduction was st°P?.ctt'
the liquid sample-collection bottles were replaced with post-run cof'
bottles.  The Tedlar gas collection bags were left in place.
                                                                    .. tnt
     The Impingers were then purged with nitrogen and fresh 0.1 N K
an additional hour to remove residual organics.  The gas collection
and post-run collection bottles were removed and the impingers wer~
drained.  The volumes of the all liquid samples and the post-run n
flow rates were measured.

                              Sample Analysis
                                                                   re ^
     A Non-Methane Organic Carbon  (NMOC) Analyzer was used to measur.e
concentration (in ppmC) of sample  in the Tedlar gas bag (non-str1pP*
fraction).  A Total Organic Carbon  (TOC) Analyzer was used to measur p
concentrations (1n ppmC) in the liquid samples (strippable fractlon/jt,fii»
liquid samples consisted of a pre-run blank, a two-hour sample, a P J.
blank, and the drainings from each  impinger.  Standard calibration * ]y#
analysis procedures were followed  on both instruments.5  The NMOC^ jftp
was used to measure the concentration of the source tank or canlste  ^
sample was Introduced as a gas.  The TOC Analyzer was used to NeaJ"tro90%) we!!?e    ti
carboxylic acids, alcohols, and low-molecular-weight alkanes.  ™*LS d"e
recoveries of aldehydes and ketones were not as good «90%), Per g|r j
loss of sample during the purging  step of TOC analysis.  In 9en!Lppal>"
classes of organic compounds that were essentially completely str K
completely non-strippable gave excellent recoveries.
     Aromatic hydrocarbons gave very low recoveries in early
example, several runs with toluene gave recoveries below 30%; sev
meta-xylene. less than 10%.  A series of tests was conducted wl tv
xylene to determine the fate of the missing material.  In the o r
impinger plumbing arrangement, the source gas came into conta$ion t^ll
following sequence of materials:   stainless steel tubing, te * ' /Oji ti
silica chips (in the glass impinger), teflon tubing, PVC tubing \   o
peristaltic pump), teflon tubing, and the Tedlar bag.  In the te
trace amount of meta-xvlene was found on the silica chips, an^j c&^ t
concentration of meta-xvlene source gas did not decrease signl* ^^n1
it was stored for a few hours in a Tedlar bag.  A series of eXP,g( ***
then conducted in which meta-xylene source gas, from a Tedlar ^, sor^
passed through the apparatus to either another bag or a charcoa
tube, and the intervening impinger, tubing, and other apparatus
progressively removed, plece-by-piece, until the recovery ex  s
the next-to-last experiment of the series, with the gas go^f   -
the source bag into the PVC tubing around the pump, thence dire
charcoal sorbent tube, the recovery of meta-xvlene was stiM
the final experiment, when source gas was pulled from the kaQ  asgd *
a charcoal tube, with no intervening tubing, the recovery         I
The PVC tubing was thus identified as the source of the
capacity for removing aromatic hydrocarbons from source gas
measured but must be very large.


                                   544
-------
       ^est recovery  (90%) of meta-xylene  1n  an  ordinary Implnger run was
   n  6c' Us*n9 VI ton tubing around the peristaltic pumps  for the non-
°f whi u e (gas) fraction.  Individual runs  with toluene  and hexane,  both
rePeat   9ave P°or recoveri'es  (Table  I) with PVC tubing,  have not been
c°ntai * w1tfl Viton tubing.  However, later  runs with  a complex mixture
      n9 koth toluene and hexane indicated no problems  with recovery of
     compounds when viton tubing was used on the  peristaltic pumps.

           recovery was 905I°  or better in all  of  the  runs with viton
            for the aldehyde and ketone mixtures.   Percent relative
       deviation (%RSD) was  less than 10%  in  all  runs.

   Us 5omPlex mixture, composed of compounds from five different classes,
   no  1n tne laboratory to  further test the  performance of the dynamic
   a?nTrfVOC^TOC systein and for comparisons  with Method 25.  The mixture
pr°P1on i equal Parts (by liquid volume) of  hexane,  toluene, ethanol,
*epar:Jaldehyde, and methyl ethyl ketone 1n  nitrogen.   Results with two
8t>e 91v  runs Us^n9 t'ie inipinger-VOC/TOC method and one run using Method 25
^y$ hvJ 1n Table n-  Tne two finger runs were conducted on different
the 5*-,, Different analysts.  Although the distribution of material between
9*Ve exr ?able and non-str1ppable fractions  1s  quite different, both runs
r*S(|lts   'ent total recovery  and with excellent precision.  The implnger
     5 were also 1n excellent agreement with Method 25  results.
               1mp1nger-VOC/TOC  method  appears to be at least as accurate
        25'  In a11 runs'  after  modifications 1n the plumbing of the
   ri"is w! Prec1s1°n was ver%y 9°od «9% relative standard deviation), even
    eton  !? <"90% recovery-  In  general,  organic compounds (I.e.. aldehydes
         ' found to cause  unresolved problems with recovery by the
        ethocl are not Hkely to  be  present as major components 1n gaseous
            source emissions.
            the research  described  In  this paper was funded by the U.S.
      r      Protection Agency,  this document has not been subjected to
      l « 5W and does not  necessarily reflect the views of the Agency.  No
   eed1nq    sernent snould be  Inferred  by  Its inclusion 1n these
                                 545
-------
References
   1.  J.  C.  Pau,  J.  E.  Knoll,  and  M.  R.  Midgett,  "Design  and LaboratjJ"
      Evaluation  of  a  Dynamic  Impinger  and  a  Continuous  Integrated
      Sampling  Cell."   Presented at  the  Symposium on  Recent  Advances
      Pollutant Monitoring  of  Ambient Air and Stationary  Sources, "a"
  2.  J. C. Pau, J.  E.  Knoll,  and  M.  R.  Midgett,  "Studies  on  the
      Impinger Sampling System - Application  to  the  Sampling  and
      of Halogenated Organics."  Presented  at the Symposium on  Recen
      Advances in  Pollutant  Monitoring of Ambient Air and  Stationary
      Sources, May 1983.
                                                                    flfld
  3.  R. K. M. Jayanty,  J. A.  Sokash, and S.  W.  Cooper,  "Laboratory
      Field Evaluation  of  Dynamic  Impinger  Sampling  System for      *
      Determination  of  Halogenated Organics from Stationary Sources-
      USEPA Contract Number  68-02-3767  (August 1985).
                                                                  t tli*
  4.  M. Jackson,  C. K.  Sokol ,  and R. K. M. Jayanty,  "Evaluation OT j
      Dynamic Impinger  Sampling System for  Determination of Halogen ^jf
      Organics from  Stationary Sources."  USEPA  Contract Number 68-
      (November  1986) .
                                                                «
  5.  Federal Register.  EPA  Method 25, 40 CFR Part 60,  Appendix Ar
      1985.
                      TABLE  I   MEAN  PERCENT RECOVERIES
    Compound (s)
propane
hexane*
toluene*
meta-xvlene
propane & toluene*
acetone, MEK & MIBK
propionaldehyde &
butyraldehyde
MeOH & EtOH
MeOH, EtOH, & iPrOH
acetic, propionic,
& butyric acids
Non-
Strippables
(Gas)
M
97.6
65.5
26.0

90.7
64.7
10.4

<5

<5
<5
<1

Strippables
(Liquid)
(%)
<5
<5


^1
<5
66.4

82.4

99.4
94.7
95.6

Total
Recovery
(%)
97.6
65.5
26.7

90.0
64.7
76.8

82.4

99.4
94.7
95.6
Re I a l j
ctanda™
De5l^°n
— -- i*^'
3.5
5-2
3.7
1 2
A •
8.9
3.0
ft 0
.1 •
A

,.
,-
*Recovery using PVC tubing
                                   546
-------
       TABLE II  MEAN PERCENT RECOVERIES WITH COMPLEX MIXTURE
          hexane, toluene, ethanol, proplonaldehyde, methyl ethyl ketone
   Inter-Run Comparisons:  Same components, runs on different days
                     Non-
                 Strippables   Strippables
                   Relative
         Total     Standard
ft)
-%*od/Run 1S)'
l[f|fi< h — ' "" " ' '
"SJsrs/Run 1 22.2
^igers/Run 2 32.8
*r 27'5
27 .3
ILIUUIU;
74.6
64.0
69.3
10.8
i\ei.uvci jr
/ V \
\ ™ /
96.8
96.8
96.8
0.0
uev i OL lun
/ Of \
\ A) /
1.6
1.5


    0mParlson with Method 25:  Same mixture,  runs  conducted  separately
                    Method
               Impinger  (average)
               Method  25
  Total Recovery
        96.8%
        93.5%
                             Depleted  Gas  Out
                             (Non-Strippables)
                                    T
                   Liquid  Level
                 Glass  Wool  Plug
                                             Stripping Liquid In
    *•
r
                                            Source Gas In
                                Liquid Out
                               (Strippables)

              FIGURE 1   COUNTER-CURRENT DYNAMIC IMPINGER
                                 547
-------
THE USE OF FRACTAL DIMENSION TO CHARACTERIZE
INDIVIDUAL AIRBORNE PARTICLES
Philip K. Hopke and Y. Debo Adewuyi
Institute for Environmental Studies
University of Illinois at Urbana-Champaign
1005 W. Western Avenue, Urbana, IL 61801
and
Gary S. Casuccio, William J. Mershon, and Richard J.  Lee
R.J. Lee Group
350 Hochberg Road, Monroeville, PA 15146
     The computer-controlled scanning electron microscope has the ability
to characterize individual particles through the fluoresced x-ray spectrum
for chemical composition data and image analysis to provide size and shape
information.  This capability has recently been enhanced by the addition
of the capacity to capture individual particle images for subsequent
digital processing.  Visual texture in an image is often an important
clue to the experienced microscopist as to the nature or origin of the
particle being studied.  The problem is then to provide a quantitative
measure of the observable texture.  The use of fractal dimension has been
investigated to provide a single number that is directly related to
observed texture.  The fractal dimension of the object can easily be
calculated by determining cumulative image properties of the particle
such as perimeter or area as a function of magnification.  Alternative
methods that may be more computationally simpler have also been explored.
The fractal dimension for a variety of particles of varying surface
characteristics have been determined using these different computational
methods.  The results of these determinations and the implications for
the use of fractal dimension in airborne particle characterization will
be presented.
                                         548
-------
aSsoc^0mPuter-Controlled Scanning  Electron Microscopy (CCSEM) with its
char ated x-ray fluorescence  analysis  system is  capable of sophisticated
Pattl i r*zati°n °f a statistically  significant number of individual
char   s from a collected particulate  matter sample.   This
alon Cterization includes elemental  analysis  from carbon to uranium and
Parti ^  ^ scanning and image  processing can  analyze an individual
often    in less than 2 seconds  although longer analysis times are
           for more complete particle analysis.   This automated
         greatly enhances  the  information that can be obtained on the
Part^ f  and chemical characteristics  of ambient or source-emitted


l'8e 0f6Cently improved data  analysis methods have been developed to make
"e ar ttle elemental composition data  obtained by CCSEM.1'2   However,
ftou fc, Cur*ently not taking  full advantage of the information available
aM tli e instrument.  In addition to the x-ray fluorescence spectrum
^ten,6 Si2e and snaPe data  available  from the on-line image analysis
Htoy'    *s now possible to  store the  particle image directly.  Major
5UoWgenients in automated particle imaging by the RJ Lee Group now
fray ^ fc"e automatic capture of  single particle images.  Thus, a 256
C"~ be Ve^"' 256 pixel by 256 pixel images of a large number of particles
   4cleasily obtained for more  detailed off-line analyses of shape and
  the S texture.  These analyses may  provide much clearer indications
  tht Particle's origin and what has  happened to it in the atmosphere.
      *ePort, one approach,  fractal  analysis, will be applied to a
 "6*PL   *maSes °f particles  with known chemical compositions in order
^ttcl0re the utility of the  fractal  dimension for characterizing
     6 texture.

^«ctai

    Tv%
^e pt e texture of a surface  is produced by its material components and
^tUitsC6ss by which it is formed.  A  technique called fractal analysis
ni  *l t>   ailed quantification  of surface texture comparable to the
 '  e*attrCeiVed by the human  eye-  Although the fractal technique is
   "sviih  lly comPlex. it: ls  simpler  than a Fourier analysis and it is
     lted for CCSEM.
(.   the f
4 Vent   Ctal dlmensi°n °f  the surface is characteristic of the
f  ch6mj ^ nature of the surface being  measured, and thus of the physics
CSibili try °f that f°rmation  process.   It is the objective of this
I  the cvty study to determine whether or not this concept is applicable
    gto  racterization of individual  particles and to their classification
       PS that can further be  examined.

i,  «• fj^ctal analysis is based on the  fact that as a surface is examined
H/' If  ^ and finer scale,  more features become apparent.  In the same
^sUtedan aggregated property of a surface such as its surface area is
J| BiagjJ a given value will be  obtained at a given level of detail.  If
^  8 to  Cation is increased,  additional surface elements can be viewed,
  4t the  t'le total surface  area measured.   Mandelbrot3 found, however,
       Property can be related to the  measurement element as follows

             - k 61'0


                                 549
-------
 where P(E) is the value of a measured property such as length, area,
 volume,  etc., E is the fundamental measurement element dimension, k'is
 a proportionality constant,  and D is the fractal dimension.  This fractal
 dimension is then characteristic of the fundamental nature of the surface
 being measured,  and thus of the physics and chemistry of that formation
 process.

      A classic example of understanding the fractal dimension is that of
 determining the  length of a boundary using a map.   If one starts with a
 large scale map  and measures,  for instance, the length of England's
 coastline,  a number is obtained.   If a map of finer scale is measured,
 a larger  number  will be determined.   Plotting the  log of the length against
 the  log of the scale yields  a  straight line.

      Mandelbrot  has demonstrated how one can use the idea of fractal
 dimensionality in generating functions that can produce extremely realistic
 looking "landscapes" using high resolution computer graphics.   The reason
 these pictures look real  is  due to their inclusion of this fractal
 dimensionality,  which provides a realistic texture.   Thus,  it  can be
 reasonably  expected that  from  appropriate  image data incorporating the
 visual texture,  a fractal number  can be derived that characterizes that
 particular  textural pattern4.

      The  fractal  dimension of  the surface  can be determined in a variety
 of ways.  For  example,  the length of the perimeter can be  determined for
 each  particle  by  summing  the number  of pixels in the edge  of the particle
 at each magnification.  The  fractal  dimension can  then be  obtained as the
 slope of  a  linear  least-squares  fit  of the  log (perimeter)  to  log
 (magnification).   Alternatively,  the fractal  dimension of  each image
 could be  calculated by  determining the surface area  in pixels  at each
 magnification  and  examining  the log  (area)  against the log  (magnification).
 In addition, the  fractal  dimension can be  determined from  the  distributions
 of the intensities  of pixels a  given distance away from each given pixel
 as described by Pentland4.

     Another approach for  the  surface  fractal dimension determination has
been  suggested by  Clarke.5  Clarke's  method is to  measure  the  surface
area  of a rectangular portion  of  an  object  by approximating the  surface
as a  series of rectangular pyramids  of increasing  base size.   First, the
pyramids have  a base of one pixel  by  one pixel with  the height being the
measured electron  intensity.   The  sum  of the  triangular sides  of the
pyramid is  the surface  area for that pixel.   Then  a  two pixel  by two
pixel pyramid  is used followed by  a  four by four pixel pyramid until the
largest 2n  size square  is  inscribed  in the  image.  The slope of  the  plot
of the log  of  the  surface  area versus  the log of the  area of the  square
 is 2-D where the slope will be negative.  Clark's  program was  written in
the C computer language.   It has been  translated into  FORTRAN  and tested
successfully on the  data  sets provided in Clarke's paper.   This  approach
appears to be  simple to use and computationally efficient and  will be the
primary method used  for this study.


Results and Discussion

     To test the ability of fractal  dimension to distinguish among several
different particle compositions, secondary  electron  images  were  obtained
for several particles each of  sodium chloride, sodium  sulfate, and ammonium
sulfate.   In addition,  images of several particles whose composition was
not provided to the data analyst were  also  taken.  The particles  analyzed

                                   550
-------
are  listed  in Table  I.   To  illustrate  the  Clarke  method,  Figure 1 shows
a 16 level  contour plot  of  the  secondary electron intensities from an image
of a sodium sulfate  particle  observed  at a magnification  of 180.   The
rectangle inscribed  in the  particle  is the area over which the fractal
dimension was determined.   The  log (surface area)  is plotted against the
log  (area of the  unit pyramid)  for the four particle types in Figures 2-
5.   The  results of the fractal  dimension analysis are also given in Table
I.   The  "unknown" compound  was
     It can be  seen  that  there  is  good agreement  of the fractal dimensions
determined for  a  series of  similar composition particles with some notable
exceptions.  Most of the  high magnification views show dimensions
significantly lower  than  that obtained at  lower magnification.   It is not
yet  clear why this result has been obtained,  but  there appears  to be a
potential systematic varation related to image magnification.   However,
for  the set of  images taken at  a magnification such that they fill
approximately 60% of the  screen area,  there is a  consistent pattern of
fractal dimension for each  chemical compound.   There is a sufficiently
small spread in fractal dimension  values that might be useful in
conjunction with  the fluoresced x-ray intensities in classifying the
particles into  types for  use in a  particle class  balance^ where additional
resolution is needed beyond that provided  by  the  elements alone.

     These results certainly provide a stimulus for further study.  It
is clear there  is a  relationship between the  measured fractal dimension
and  the texture observed  in the images .  Further  work is needed to
determine a method for routinely obtaining reliable fractal dimensions
from secondary  electron images .


Acknowledgements

     The work at  the University of Illinois was supported in part by the
National Science  Foundation under  Grant ATM 85-20533.
References

1.   D.S. Kim and P.K. Hopke,  "The  classification of individual particles
    based on computer-controlled scanning  electron microscopy data,"
    Aerosol Sci. Technol.  (in press,  1988).

2.   D.S. Kim and P.K. Hopke,  "Source  apportionment of the  El  Paso aerosol
    by particle class balance analysis," Aerosol Sci.  Technol.  (in
    press, 1988).

3.   B. Mandelbrot, The Fractal  Geometry of Nature,  W.H.  Freeman and Co.,
    San Francisco, CA (1982).

4.   A.P. Pentland, "Fractal-based  description  of natural scenes," SRI
    Technical Note No. 280, SRI International, Menlo Park,  CA (1984).

5.   K.C. Clarke, "Computation of the  fractal dimension of  topographic
    surfaces using the triangular  prism surface  area method," Computers
    & Geosci. 12: 713-722  (1986).
                                  551
-------
           Table I.  Particles Analyzed for Fractal Dimension
No.
1
1
1
2
2
3

3
4
5
5
6
6
7
7
8
8
9
9
10
10
11
11
12
12
Particle
Type
Na2S04
Na2S04
Na2S04
Na2S04
Na2S04
Na2S04

Na2S04
(NH4)2S04
(NH4)2S04
(NH4)2S04
(NH4)2S04
(NH4)2S04
NaCl
NaCl
NaCl
NaCl
NaCl
NaCl
Unknown 1
Unknown 1
Unknown 2
Unknown 2
Unknown 3
Unknown 3
Magnification
180
1000
5000
250
1000
100

1000
150
330
3500
270
1000
330
1000
330
2500
350
3000
275
8500
1600
7000
2000
7000
Fractal
Dimension
2.41
2.44
2.33
2.42
2.42
2.36

2.43
2.28
2.23
2.19
2.26
2.20
2.27
2.27
2.29
2.13
2.27
2.15
2.36
2.25
2.38
2.30
2.37
2.29
Uncertainty
+- 0.01
+- -0.01
+- 0.02
+- 0.01
+- 0.01
+- 0.02

+- 0.01
+- 0.02
+- 0.02
+- 0.02
+- 0.02
+- 0.02
+- 0.01
+- 0.01
+- 0.01
+- 0.01
+- 0.02
+- 0.01
+- 0.02
+- 0.02
+- 0.04
+- 0.03
+- 0.03
+- 0.03
	 	
Average
Value
___ 	 	





-d
2A.12
.41*



-b
2.23




9 27^





370d
-^~^
a) Excludes Particle 1 at SOOOx and Particle 3 at lOOx.
b) Excludes Particle 5 at 3500x and Particle 6 at lOOOx.
c) Excludes Particle 8 at 2500x and Particle 9 at 3000x.
d) Excludes Particle 10 at 8500x and Particles 11 and 12 at 7000X-
                                   552
-------
w™ich
     >
         particle showing inscribed rectangle within
surface area is determined.
              553
-------
           024

                          Log  (Unit Pyramid)


Figure  2. Plot of log (surface area) against  log (unit pyramid)
         three Na2SC>4 particles at each magnification.
     4>
     U
     *
     «M
     fa
     0
          024

                         Log  (Unit  Pynmid)

Figure  3. Plot of log (surface area) against  log (unit pyram
         three (NH/^SO^ particles at each magnification

                                 554
id)
   fo*
-------
                     2               4

                    Log  (Unit Pyramid)
   Plot of log (surface area)  against log (unit pyramid)  for the

   three NaCl particles at each magnification.
I
I
,
                     2               4               6

                   Log  (Unit  Pyramid)

    /•ot of log (surface area)  against log (unit pyramid) for the
    thr
      ee (NH4)HS04 particles at each magnification.


                           555
-------
STATISTICAL DETECTION OF CHANGES IN
AMBIENT LEVELS OF TOXIC AIR POLLUTANTS
Mithra Moezzi
Thomas 3. Permutt
Alison K. Pollack

Systems Applications, Inc.
101 Lucas Valley Road
San Rafael, CA 94903
In 1985 the California Air Resources Board collected biweekly measuremen
                                                                i  11 rtn<* *  f
ambient atmospheric concentrations of benzene and other aromatic and  ,y^pj$
hydrocarbon species at 21 sites in California. Collocated measurements o   . ^
criteria pollutants were also collected. The relationships between the leve. nS
gases are examined.  For most sites, the data show strong positive correia  ^
among the daily average levels of the several aromatic species and betwe  ^ ^tr
levels of the aromatic species and the levels of carbon monoxide and oxide  tion°
gen. We show how these strong correlations may be used to aid in the de
changes in concentrations of organic species using as an example the Pr .    c<
detecting changes in benzene concentrations under a hypothetical erni$sl    ^
program. At the monitoring frquency used for the California data kase ?,orrlia *|
benzene concentration distributions typical of those observed at the Can    ^/e
toring stations, a simple classical statistical test for change in mean ben ^  ,
from one year to the next will more likely than not fail to detect a stati
significant decrease in benzene concentrations when in fact concentrati  ^
been reduced by one-fourth.  By incorporating information provided by   ^
monoxide measurements collected along with the benzene measuremen  *
changes of detecting such a change can be substantially increased.  Nu
examples are provided.
                                    556
-------
 Auction
  El
   evated atmospheric concentrations of some volatile organic compounds (VOC)
 ijjj,  e Potentially toxic and carincogenic effects on the humans exposed to them.
  an-8 Contro1 programs may be instituted to control the concentrations of particular
    l{fe* To assess the effectiveness of a VOC control program, a statistical test can
        using data collected before and after the control program in order to assess
    °r not atmospheric concentrations have actually decreased.  The statistical test
  itL!° ^at tne chances are small of deciding that there has been a change in con-
       f in fact no such change has occurred. However, there is no a priori guaran-
       statistical test will usually correctly detect a change in concentrations if in
       rations have changed.  The probability that the statistical test will correctly
 v^5- fences in concentrations from one period to the next is called the power of
J^*h   *S essential that the approximate power of the test be known in order to
^h.-  '"er the failure of the test to detect a difference in concentrations between
   .   l°ds implies that no change has occurred.  If the power of a test is low, the
      °* the test will likely be that no change has occurred, even if the level of con-
      actually changed substantially.

    ' *nvestigated and compared the power of various statistical tests to detect
    i twtrends *n ambient concentrations of VOCs.  In carrying out this study we
    th   ^alifornia air toxics data set to characterize distributions of organic
     ls data set is described in Section 2.  In Section 3 we discuss the power of the
     ,    *ple means test. Then we show how the power of the means test can be
      y taking into consideration measurements of a concomitant variable, such as
      ration of another VOC or of  carbon dioxide or nitrogen  dioxide. Next we dis-
         °f a nonparametric test, the Wilcoxon rank-sum test, and the power of
      c test for temporal trend. In Section 4 we summarize the  results.


     llfornia Air Toxics Data Base

    ^'eS nf
       ' °i ambient air collected at  19 locations in California in 1985 were
        12 gaseous aromatic and halogenated aliphatic hydrocarbons (volatile
       lp°unds, or VOCs). These compounds, which we will refer to as "toxic gases,"
       '  Oluene, o-xylene, m/p-xylene, methylene chloride, trichloroethylene,
      ^., etnyl chloroform, ethylene dibromide, ethylene dichloride, perchloroethy-
      2^?n tetrachloride. The data,  provided to us by the California Air Resources
     Jnt  Uf average concentrations in parts per billion volume.  Collocated hourly
    °f nitr of carbon monoxide, sulfur  dioxide, nitrogen dioxide, and ozone, total
             and sometimes methane, non-methane hydrocarbons, and total
             also taken.  Ambient air for the analysis of the toxic gases was
               rate over a 24-hour period. The starting hour of the  24-hour
      eu? .Varies by location; it is generally not  midnight. Samples were obtained
          Weekly.  The data for most locations span the entire year.
      sitfts
      c^ed •  ^ocated in relatively heavily polluted urban areas.  Thirteen of the 19
      ^»ch  m Coastal air basins. Five are located in the San Francisco Bay Area
      fal Q10^1 Concord, San  Jose, Fremont, and San Francisco); two are in the
          °ast  Air Basin (Santa Barbara and Simi Valley); four are in the South
                                   557
-------
                                                                              jjjll
Coast Air Basin (Los Angeles, Long Beach, Upland, and Rubidoux); and two are in ®  |,
Diego Air Basin (Chula Vista, El Cajon). Five sites are located in inland agriculture
leys: one in the Sacramento Air Basin (Citrus Heights), and four in the  the San
Valley Air Basin (Fresno, Bakersfield, Stockton, and Modesto).  The last site is
the Southeast Desert Air Basin (Lancaster),
3  Technical Analysis

                                    Means Test

     Suppose that one wants to compare the mean concentration level of a p°^u  g\$
before the inception of an emissions control regime with the mean  level after the
is in full effect. Assume that one has nj independent and identically distributed    ^
surements of concentration, Y.,,  i  = 1, ... n. ,  collected before the prograrT1 . f {,
and a second set of n2 independent and identically distributed measurements, i[2'0i$
... n2, collected after the controls have been instituted. A means test can be us* '   jf
whether or not  the mean concentration decreased after the controls were instit
the sample size is small, then the underlying distribution of concentrations mu   $
approximately normal and the variances approximately equal for the means test
appropriate.

     Assume that the requirement concerning normality holds, and that the
tions have equal and known variance  o  =0=0  .  Define the hypotheses

                                      HQ:  y,  <  u,


                                     Halt:   U1 > *2

     For a test at level a = 0.05, the null hypothesis of no reduction  in mean
concentration will be rejected in favor of the alternative that mean concentra
decreased if
The power of the test is
                                                                     ii\&
     To generalize, let nj = n^ = n and characterize variability by the coen ^
variation, CV = o/^.  By substitution into the above expression, the power o
be written
-------
Pw   's independent of the mean for a proportional change in concentration. We com-
fy-  . ^e power of the means test for a sample size and levels of variability charac-
 e^ lc of those for the VOC monitoring stations included in the California Air
  "'lrces data base for 1985.

">«an   Ure * snows tne power of the means test as a function of the true decrease in
   Concentration for several levels of variability, with the level of the test fixed at a
       a samP^e s'ze °f 20 from each population. Observed levels of variability in the
       data are listed  in Table I, which lists by species the median, minimum, and
       site-specific coefficients of variation observed.

    •  median coefficient of variation for benzene is 0.54. Figure 1 shows that for a
        of variation 0.50, a - 0.05, n = 20, and a 10 percent reduction in true mean
            the probability that the test will provide sufficient evidence of a decrease
    .      Even if the mean concentration were reduced by 40 percent, there is a
     itv of 0.20 that the test will not detect a statistically significant decrease. At
           and for some species the situation is even worse.  For example, the median
          ^ ^e coefficient °f variation for m,p-xylene is 0.73. At this level of vari-
          is little better tnan a ^O percent chance of detecting a 40 percent decrease
     concentration. For benzene, the power of the test for a 10 percent change in
  Ut  ean concentration ranges from about 0.09 for the station with the highest coef f i-
V*riat Variation (1-04) to about 0.23 for the station with the lowest coefficient of
'^test"1^'3^"  For a 20 Percent decrease in mean benzene concentration,  the  power of
"•5(J{0 ranges from 0.15 for the station with the most variable  measurements to about
      '"6 station with the least variable measurements.
     n way to increase the power of the test is to increase the sample size and thus
    u   friability of the sampling distribution.  Figure 2 shows power curves com-
    Po  • the assumPtions discussed previously, but with 50 observations per sample,
  ^itic ndinS to approximately weekly sampling, instead of 20 as assumed for Figure I.
 !Sges^ased sample size improves the power of the test considerably, but for moderate
 °f eXft ln concentration and typical levels of variability, the power is still inadequate.
 %0n j ple> the probability of detecting a 10 percent decrease in mean benzene concen-
  ^j8 ' ab°ut 0.25 when the coefficient of variation  is about 0.50, compared to a
       K of °'l° when only 20 observations per period are considered.  Table II sum-
      •   p°wer of the means test for three levels of variability and two different
                              Covariance Analysis

      th    Way to increase the Power of the test is through covariance analysis, which
      Or e ef Active variability of the measurements of interest by removing the pro-
     iUt that variability that can be related to the level of a concomitant variable.
       o   -SUCh tests depends on the type of control program and on the strength of
        nships between the subject and concomitant variables.

    ^ Corilier analyses of tne California air toxics data we found strong linear relation-
 V^ts Centration levels among the aromatic pollutants and between the aromatic
   '    ^ Sorne criteria pollutants, particularly carbon monoxide and nitrogen dio-
          Correlation between benzene and carbon monoxide measurements is above
        r  Stations-  For ail stations combined the correlation between benzene and
        °Xlde is 0.74, based on 397 pairs of observations.  Presumably this strong
                                   559
-------
                                             J ffjffl
                                          itte0
association exists because both benzene and carbon monoxide are primarily
gasoline.

      A control program may be designed that does not change the relationship &et j
the relative emission rates of benzene and carbon monoxide from gasoline, but aisl.-^
targets industrial emissions of benzene or controls total emissions from gasoline-    .,
benzene and carbon monoxide concentrations are well correlated, carbon monoxid
surements can be incorporated into the means test to yield a more powerful test.   ^
Assume that the slope of the regression line for benzene on carbon monoxide rem   .^
same before and after the controls are implemented. The validity of such an a
can be tested. Then a statistical test can be conducted for changes in benzene
trations adjusting for carbon monoxide concentrations.
                                                                         ssu
      Let E(Y|X) = $oi + 0nx, i = 1, 2, where i = 1 refers to the population before
control program starts, and i = 2 refers to the population after controls have bee
implemented.

      As  discussed, we assume that B..  = 6 2 ; thus the test for a change in &e
test for change in the intercept term of the regression model.  The hypotheses a
                                                                           t

                                      HQ:  BQ1 * 6Q2


                                    Halt:   S01 > B02
The variance of the sampling distribution of  (eQ_ - 601 )   is now
Var(Y|X  =  0)  = 2(1
                                                     r
                                                         Var(Y)
where r  denotes the coefficient of determination.
Reject the null hypothesis of no decrease if
                                   2(1 - rp  Var(Y)
                                       n
                                                     i <
Since SQ2 - 6Q1  =  y~ - p.,  the power of this test is
                                 *I-z  -
                                     a
                                             ^2^1  " 1
                                          cv
                                              2(1  - r*)
where CV  =  — as before.
            y1
     Figure 3 shows power curves for the test at the 0.05 significance *e
of variation  0.50, and sample size for each population of 20.  Curves are

                                       560
-------
j
j.. s°f association between the dependent and concomitant variable (r  = 0, 0.30, 0.60,
W •    ^e ^unct'on f°r the 2ero coefficient of determination is identical to the
QI  l°n for the simple two-sample means test for the given level of the test, coefficient
   lation, and sample sizes.

     a tyPical site» sav Rubidoux (coefficient of variation for benzene 0.52; correla-
    tw«en benzene and carbon monoxide 0.77), for sample sizes of 20,  the estimated
DW   °f the two-sample means test for changes in mean benzene level using carbon
   *1(te as a covariate is 0.45, compared to 0.15 for the means test conducted without
   formation about carbon monoxide.

 ijj.  ternatively, consider a control program that prospectively reduces the relative
    11 rates of benzene and carbon monoxide.  Then a statistical test can be conducted
     W^etner ambient benzene concentrations relative to carbon monoxide concen-
  de   &Ve cnanged by testing whether the slope of the regression line  has changed.  In
"*v!*lopment of tne analytical expression for the power of the test we assume that
         °f  the indePendent variable (carbon monoxide in the example described) and
      cient of determination for the regression are the same for both periods.

     e hypotheses for a test for change in slope is
                         HQ:  sn  < B12
                       Halt:  011  > 612
                           i

     S    that the variance of the independent variable x is the same in each
         null hypothesis is rejected if
                                A      A

                                812 ~Bl1
                          [2(1  -r2) Var(Y)]
                             n Var (x)    j
 w ^  ^Hcft r\             t    *
   \\ to LVar(Y)/Var(X)P = e. Jr, the numerator and denominator can be divided
       °btain the following expression for the power of the test for change in slopes:

           shows the power for the test for change in the slope of the regression line
  ij^ifc,^ of the percentage reduction in slope for coefficient of variation 0.50 for
  C^Weh* deternr>ination 0.3, 0.6, and 0.9 and a sample size of 20.  For a monitoring
      f*ct Sh°WS that the Probability of  detecting a decrease in slope is about 0.8
         the slope of the regression line has decreased by 30 percent.
  L T^,.    |
  M,*fiw .I8h association between the concentrations of some VOCs and criteria pollu-
               California data suggests that covariance analysis may provide an
              of increasing the power of statistical tests to detect changes in mean
                                    561
^QfaLso
 V!> the
     ^ethod0
-------
                                                                             lySJi
concentrations. An important difference between the means test and covariance an
test is that the former tests for absolute change in concentration while the latter te   .j
for changes in concentrations relative to the concentration of another pollutant. ™
adverse meteorology results in elevated concentrations of a VOC, the means test m '
show that concentrations have  increased while a covariance analysis test may shoW
the covariate VOC concentrations are unchanged. The assumption that the P<"e~c°nfter
relationship between the dependent variable and the covariate remains unchanged *•
controls are instituted is a strong one that  may not be valid for a particular progra
                              Wilcoxon Rank-Sum Test
                                                                             tj0n
     Estimates of the power of the means test do not depend strongly on the assu Y {
of normality, but for non-normal distributions adjusting for the Wilcoxon rank-sum^
may be more powerful than the means test, particularly if the distribution is nf  .
tailed. Our analyses of hourly  VOC concentration data from a monitoring station    c
Baton Rouge, Louisiana indicates that VOC concentration measurements may be
terized by this type of distribution.

     The procedure for conducting a rank-sum test is as follows.  Combine the
from each population and rank  the observations from lowest to highest value; sur\:stri'
ranks of the observations in one of  the samples; then compare that sum with the
bution of sums expected if samples were identical. If the sum of ranks for the g
sample is unusually low, then conclude that concentrations in that one populate  Q{
lower than those in the other population.  No assumptions regarding the distribu
concentrations are required.

     We used a bootstrap procedure to estimate the power of the Wilcoxon ran   Q\
test for detecting changes in ambient levels of benzene.  The asymptotic prop6  . ^
the bootstrap with respect to the rank-sum test are not obvious, but we think 1    $,
provide reasonable estimates.  CARB benzene data were used as subpopulations ^
bootstrap. Simulations were completed for eight alternative levels of concentr   jjp
reduction for each of the 19 sites, with 300 bootstrap samples drawn for each    J0
each repetition, two random samples were drawn with replacement from the i    $ ^
observations available at a site.  The samples drawn for each site were equal *   ^
                                                                      * to»
that of the original sample. One of the two samples was scaled by a constant
reduction in ambient benzene level; the other sample was left unchanged.  ^e
samples were then compared to each other using the Wilcoxon rank-sum test.   ^
hypothesis of identical populations  was rejected in favor of the alternative tn
trations in the second population were lower if the observed significance leve   g
than 0.05. The proportion of cases in which the null hypothesis was rejected g
estimated power of the test.
                                                                      •
     Figure 5 shows the results of the simulation.  Alphabetic plot symbols m
different monitoring stations.  Numeric plot symbols indicate that the points   j^i
stations overlap.  For each of these stations, the estimated power of the ran
higher than that of the means test. The notably low power of the rank-sum
Lancaster station {plot symbol P) is a result of small sample size.  Only  four
of benzene were available for the station; the bootstrap in this case may no
priate, but the results are included for completeness. Thus, the rank-sum te
data appears to have power sometimes better than, and not significantly
simple  means test.
                                       562
-------
                               Temporal Trend

L  Depending on the control program, it may be appropriate to test for a gradual trend
kri riCentration rather than a difference in concentration between only two time
N'cat   ^ests for trends in the ambient levels of  most volatile organic species are com-
Cjjj, ecl Because the pattern of  concentrations may be strongly seasonal.  The problem
s?a. e Deviated by removing the seasonal trend or by using a test unaffected by the
     cVcte, such as the seasonal Kendall test; however, it may be difficult to
  ar   ak°ut the power of such tests, As in the two-sample tests, year-to-year
4^  l°ns in meteorology may affect pollutant concentrations and thus confound tests
% + at Assuring control program effectiveness. Furthermore, it may not be reason-
^t° assume that anY particular control program would result in constant or even pro-
      change across seasons.

   ,,  test for trends in annual mean concentration avoids the problem of seasonal
   ~ent and is not based on the assumption that changes would occur smoothly over
              , such a test is less cost-effective for detecting an overall change in
^f0' per unit sample than is a program in which samples are collected only in the  period
    and the period after the presumed change is fully realized.
   To t
      test for a linear trend in annual mean, assume the regression relationship

                            Y = 8- + 0  X +  e,    e - N{0,  a2),
                                 VU     1
   «y
      ls the annual means and X is the year (0, 1, 2, and so on).

WjJ*test *or a downward linear trend in annual mean concentration, define the
   tr*ses

                          HQ:  B,  * 0

W                    "alt:  61  < °
  ™Q{ ii
        n
-------
                                                                             20
the means test and is shown for purposes of comparison. At the 0.05 level, a linear
percent overall decrease over a period of four years has a probability of 0.37 of
detected as a decrease.
4  Summary and Conclusions

     For a typical California Air Resources Board monitoring station, about 20 rfle
surements of 24-hour average concentrations for in 1985 are available for each of    j
species. At these monitoring frequencies and the distributional characteristics obs
for the California stations, use of the straightforward means test to measure a dec
in VOC concentration provides inadequate power to detect a decrease unless the c
in concentration is at  least 40 percent. In some cases the Wilcoxon rank-sum test    jj
provide more power than the means test, but in general the power of the Wilcoxon    ,
still low for detecting moderate decreases in concentrations.  Larger  samples, obta ^
by increasing monitoring frequency, improve the power of both tests,  but monitor1 B^
expensive.  In some cases collocated measurements of an air pollutant that are #e
related with the level of the subject VOC may be available. If so, these measu
may be used in conjunction with measurements of the subject VOC to potentially
stantially increase the power of the means test.
Acknowledgements

     This study was performed under contract to the Monitoring and Data Analy ^
Division of EPA's Office of Air Quality Planning and Standards.  The authors ack
ledge the guidance of Tom Curran, Chief of the Data Analysis Section.


References

1.   T. 3.  Permutt, M. Moezzi, A. B. Hudischewskyj and C. S. Burton, "Statist^
     sis of Concentrations of Toxic Air Pollutants in California and Louisiana*
     Applications, Inc., San Rafael, California (SYSAPP-87/131), 1987.
                                                                            gfjf
2.   M. Moezzi, "Prediction of Concentrations of Air Toxic Gases from Measu
     of Criteria Pollutants," Systems Applications, Inc., San Rafael, California
     (SYSAPP-87/012), 1987.
                                                                            t,\t

3.   A. K. Pollack, T. J, Permutt, M. Moezzi and R. G. Johnson, "Statistical Pr
     of Hourly Concentrations of Volatile Organic Compounds at Baton
     Louisiana," Systems Applications, Inc., San Rafael, California (SYSAPP"
     1987.

                                                                  „
-------
   -  Coefficients of variation  for measurements of  24-hour  concen-
ts  of  volatile organic  compounds.
Number of
Compound Monitoring Statioi
Sne
TO),
iuene
Vnl
^th^lene
Chi/1 °hloride
Coform
Cifbyl Chloroform
Etlty. tetrachloride
Ethyi e dibromide
PW*e dichloride
Moh °6thylene
^oethylene
19
10
10
10
19
19
19
19
19
10
19
19
Site-Specific Coefficient
of Variation
is Minimum
0.39
0.17
0.26
0.49
0.57
0.30
0.53
0.07
0.0
0.0
0.26
0.05
Median
0.54
0.47
0.60
0.72
0.86
0.70
0.73
0.28
0.91
0.22
0.47
0.44
Maximum
1.05
0.83
1.23
0.93
1.97
3.00
3.13
0.62
2.18
0.94
1.23
0.70
•E
  II.
                power of the means test for detecting
        ambient concentration levels between two
Uta n^ Perioc's as a function of sample size and
  nt concentration coefficient of variation.
e Coefficient
0.
Sample
20
0.27
0.67
0.93
0.99
1.00
1.00
1.00
30
Size*
50
0.51
0.95
1.00
1.00
1.00
1.00
1.00
0.
Sample
20
0.15
0.34
0.59
0.80
0.93
0.98
0.99
of Variation
50
Size*
50
0.26
0.64
0.91
0.99
1.00
1.00
1.00
1.
Sample
20
0.12
0.22
0.38
0.55
0.72
0.85
0.97
00
Size*
50
0.18
0.41
0.69
0.89
0.97
1.00
1.00
    of observations per period.
                              565
-------
1,00
0.90 -
0.10
0.00,
            10
20        30        «0       50

  Porcsntoa" Reduction in Mean Concentration
                                                                  Jo.oo
FIGURE 1.  Estimated power of two-sample means test  as a
of percentage  reduction  in mean concentration, assuming 20
observations per sample.
                                                                   :tion
 1.00
 0.90
 0.00
                      20        30       *0       50

                        P«rc«nlog» Reduction in M«on Conc«nlrolion
  FIGURE 2.   Estimated power of  two-sample means  test as  ^
  of percentage reduction in mean concentration,  assuming
  observations per sample.
                                  566
-------
                                                                1.00
                                                               - 0.90
                                                               - 0.80
                                                               - 0.70
                                                               - 0.60
0.00 L
           1O
              2O        30       40       50
                Percentage Reduction in M*on Concenlrotion
                                                      6O
f      3.  Estimated power of analysis of covariance  test as a
 fiction of percentage reduction  in mean concentration,  assuming
 u Paired observations per sample and coefficient of  variation
 >5°»  for coefficients of determination 0.0, 0.3, 0.6,  and 0.9.
           10
                    20        30        40       SO
                      Percentage Reduction in Mean Concentration
                                                       60
 0 3
  '
  4.  Estimated power of the  test for change  in slope of
      line as  a function of percentage reduction in mean
 -—auions, assuming 20 paired observations per sample and
 •cient of variation 0.50, for coefficients of determination
°-6, and 0.9.             567
-------
 0.00
            10
                     20
 30        *O        50

g* Reduction in Concentration
FIGURE  5.  Estimated power  of the Wilcoxon rank-sum test for, e
California Air Resources Board monitoring stations.  Solid  I1
indicates power of the two-sample means test for coefficient
of variation 0.50  and 20 observations  per sample.
   1.00
  0.90 •
  0.10
  0.00
              10
                       20        30        40        50

                     Total Percento?* Reduction in Concentration Over
                     Total Percentage Reduction in Concentration Over Pmioo           c
  FIGURE  6.   Estimated power of trend test as a function o
  percentage reduction in mean concentration,  assuming
  observations per  year and coefficient of variation 0.

                                 568
-------
>1SD*TRIAL TOXIC GAS STORAGE FACILITY -
  ^RSION STUDY
      R. Murray, CCM
  g '.Hoffnagle, CCM
  Jvironmental Consultants, Inc.
        ord* CT  06108
    289-8631
     "^nufacturing facility  set  in  a  flat  terrain,  suburban  area
   w a Si9nificant amount  of phosgene on site.  Although  the  storage
    *s very secure with full double containment, both plant personnel
        Community  were   concerned  about  the  consequences  of   an
          release.    An  atmospheric  tracer  study  was  designed  and
        to <3uantify  fenceline  concentrations  arising  from  continuous
     »h*such  as  Pi?6  or  flange failures)  from the phosgene  storage
  ^U     field measurement program  emphasized quantification of  the
      C°ncentra*;ion  and  the cross-wind structure of the  plume.   Over
         Concentration observations  were  made during  different  plume
         a"d meteorological conditions.   Observed  plume patterns were
   8$v  to Predictions  made  with the EPA's  ISC  model.   ISC  moderately
      **    overpredicted   observed   maximum   concentrations  at  the
          d underpredicted  the  plume widths.   For  emergency response
         ese results have  significant implications:  The distance for
   the e.plume is expected to be above  safety thresholds is reduced,
 * dHScWlc*th of the  zone  is increased.  These results  and comparisons
      ussed in this  paper.
                               569
-------
Introduction

    Toxic and hazardous  chemical  storage areas  are  frequently Par-f0(i
large  industrial  complexes.   The dispersion  of plumes  resulting   ^
the  release of  these  chemicals  within  building  complexes  wil   j,
strongly  influenced  by the  buildings  themselves which will  alter    g
                                                                      j,

the transport  flow field  and the  turbulence.   As  a  result of  t
source   related   phenomena,   traditional  dispersion   models  may
adequately   characterize   plume   trajectories   and   dispersion.
recognition  of  this  possibility,  an  atmospheric  tracer
undertaken  at  a  phosgene  storage  facility  associated with  a
manufacturing  plant.   The study  was  not  intended  to be an
dispersion experiment, but was  designed to be  indicative of
phenomena at the plant.  The objectives of the study were to:
    •  Provide  measurements  of   plume  direction   and  downw
       concentrations under a variety of meteorological conditions-

    •  Provide a  site  specific  set of dispersion  data  that  can
       used  to evaluate  emergency  response  models  and calibrat
       those models if necessary.
Experimental Method

    The  tracer  study  was  designed  to  simulate  continuous
emissions  (duration  of  fifteen  minutes  or  longer)  from  the
facility,  such as  those  caused  by  a  flange  or  valve
experimental  design  was  prepared  calling  for  four  tests      O
daytime and nighttime  releases and with wind  directions  from b°
the plant complex and over the adjacent forested area.
                                                                   r30'
    A  tracer  release system  was  located  next to the  phosgene     jfc
area,  and release  rates  were set using  a calibrated  rotame^ 'e'*
    Twenty-seven  tracer  samplers  were  deployed  on a  foreca  gou^
along  a  single arc  at the minimum fenceline  distance  from  tn«  ^i***,
675 meters (see Figure  1).  The sampler spacing  (six degrees <> ^fcgj &
for unstable  to neutral  tests and three  degrees  for stable fc   ^ t^
designed  to provide  high resolution of the  cross-wind  struct° c& * .$
plume  and to  resolve  the maximum  concentrations.   Prior  to e^te^tf
the Test Director/Meteorologist prepared a  site  specific wind ^jd  M
forecast  to  determine  which potential  sampler  locations     ^ ,
instrumented for  each  test.   Additional  samplers were  deploy  f \ .
the   building  complex  and  at   the  ends   of  a  500   m   ^.  $
warehouse/shipping building adjacent to the  phosgene storage ^  ^tf
tracer samplers were  programmed to collect nine, fifteen-minu
                                   570
-------
Sp,
 6  samples  per test.   Subsequent  to each  test,  the sample  syringes
fa ?  returned  to   a  laboratory  established  near  the   manufacturing
aij..1   y f°r concentration analysis using gas chroma tography.   To  avoid
ua][. Possible  cross-contamination  between  tests,  clean  syringes  were
 ea for each test.

10  7?  anemometry  set was  deployed in  a clear  area  near the  release
the  1On and a**  ^e same height  as  the release point  (two meters  above
mn  Around) .  Wind  speed and  direction data were  recorded as  fifteen
pet:    averages  corresponding  to  each  fifteen minute   SFs   sampling
              site   cloud   cover   and  obstructions   to  visibility
            were also collected so  that Turner stability classes  could
(ISc\
    *ter the  meteorological  and  concentration data  were  tabulated,
         Per>iod was modeled  using the EPA's Industrial  Source  Complex
    - model.  ISC  is designed  to simulate concentrations  arising  from
    ions within  building  complexes.
   TVi
    n®  first   test  was   conducted  mid-day  with  a  heavy   overcast
    \lri<3 in  Turner  'D1  or  neutral  stability  throughout the  period.
    '*nds blew  away from  the building complex  and  over  the  forested
     *th a systematic  northwest  to northeast clockwise wind  direction
     fluring the test.  Figure 2  presents a comparison of  the  observed
   h>tedicted plume widths  and maximum concentrations for  each fifteen
      sampling block.  Plume widths  have been  defined  as the  angular
      e between the samplers where the plume  concentration falls below
   _. cent  of the  maximum  centerline  concentration.  In the event that
         centerline  was  observed near the  edge of the tracer  grid,  a
   ,„       plume  distribution  was assumed, and  the plume  width  was
  'cen,as  double the  angular distance  from the maximum  value  to  ten
   6 iHadCOnCentration value  tnat  was observed.   Predictions using  ISC
  1 86tf •  f°r  actual sampler locations using the  observed wind speeds
       lnfT  the  plume  centerline to the  azimuth of  the observed maximum
                   building  downwash algorithm  in ISC was invoked to
           the  influence of the warehouse/shipping  facility.
             of Figure  2  reveals  that  the monitored plume was  much
        than   the  predicted   plume   and  that   ISC   significantly
             the maximum  concentration values.  There are  probably at
          factors  contributing  to  the  much  broader  than  predicted
               smoke  tests and  concentrations  monitored at  the  tracer
        *t the  corners  of  the  warehouse/shipping  building  (samplers
   -*«iq< B' on pi9ure 1)  show  that  the plume  became  entrained  in the
 L*  wa«.S  Wa^6 zone.  Even though  the  phosgene  storage  area/release
         about  thirty  meters   from   the warehouse  and  the wind  was
            from the  warehouse, the  plume became  trapped in an  eddy
         tfie  warehouse building and  was carried along the face  of the
          The resulting  plume was  broadly dispersed  in  the wake  zone
         lts  downwind trajectory from an effective  volume source.   The
      iin  
-------
     There  are  several  implications  of   this  result   for
                                                                     3i*w
 warehouse,  the  entire warehouse  area  is  subjected  to the  plume
 safety precautions must be  followed  in the warehouse during a  relea!ue
 The second implication  is  that a  much broader  area is subject to   ,
 plume.   The ten percent of  centerline  concentration limits is  as  W
 as 80  degrees,  whereas  modeling using ISC would indicate plume wi<*
 of only 20  degrees.   Finally,  the maximum  concentrations are much  1°  t
 than the modeled values.  This  indicates  that the potential hazards
 off-site locations may be  overstated  by the model.

     Figure  3  presents the results  for  the  second test which took  P   ^
 under  clear skies  at mid-day with  a  plume  trajectory over the  buil
 complex.   The Turner stability was  'A'  or  very unstable throughout ^
 test.   The  plume direction  remained  steady throughout  this  test.     9
 observed versus predicted plume widths and maximum concentrations    ^
 the closest for this test of  any  in the test  series.   Generally'    .$
 the plume widths and maximum  concentrations were  overpredicted  but
 seriously than  in any other  test.

                                                                 • t-h ^
     Figure  4  shows   the  results  of  the first nighttime  test  wl   wef6
 plume  once  again over the  building  complex.   Although conditions   ^\
 very stable  (light   winds with  clear  skies  resulting in  a  Turner ^
 stability throughout the  test period),  the  observed plume  wi^   •$$
 nearly   equal   to  the  daytime  very unstable  condition  plume ^ ^
 observed in the second  test.   The   ISC model again  overpredicte   j,,
 maximum   concentrations    and   underpredicted    the   plume      $9
 Observations   of  smoke   releases   revealed  that   once   againt t}i«
 warehouse/shipping  building acted  as  a flow obstruction  and  tn*   ^
 plume dispersion was enhanced  as it  cleared  this  first obstacle *
 path.

    A  second  interesting phenomenon  occurred during  this  test
 the  plume  passed  the building  complex,  it  was  detected  by
 located in  open areas along the plant's  parking lots  and by ^
 within   the  forest  canopy.    Steep  concentration  gradients
 during  the  test between the  "in-canopy"   samplers  and  those
 with    clear    exposures.     The    in-canopy   samplers   had
 concentrations.  Apparently  the tracer  material  that entered the   ^.
 became  cut  off  from  the  ventilating  effects  of the  light  w*nhe  oP8J
 evening,  and was not diluted as  rapidly as tracer material in  A\tf$
 fields.   Concentration gradients  of a  factor of  ten between  &
 in-canopy versus out-of-canopy samplers  resulted.                     .
    Figure  5  gives the  results for the  final  test which was co
under  very  stable conditions,  Turner   'F'  stability,  with  * ati ^
trajectory  away from the  building complex.   The plume  width
fairly narrow  for  this  test but widened  as the  plume  trajectory^
the   sampler   array   from   north  to   east-northeast.    ^
consistently  underpredicted  the  plume  width   and  overpredi
maximum  concentration.   Again  during  this  test  the warehouse
building's  wake   trapped   the   plume.   As  the  measured  ce
direction  of  the  plume shifted  from 12  degrees to 33  degce^tio(I  p
this test,  the concentration at  the  sampler  located  at          '
(see Figure 1)  climbed from  54  ppt to over  9700  ppt.
migration  of  the plume  along the building face  is  very
small changes in wind direction.


                                   572
-------
    ry
   Plumes released near ground  level  within or adjacent  to  a building
IM lex wil1 be Stron9ly influenced by the flow patterns and turbulence
r [JUced by  the  buildings.   Both  the trajectory of the  plume and  the
do   °^ dispersion of  the  plume  will be strongly influenced.   The plume
 "^wash/building  cavity zone correction  algorithm included in the  ISC
Ov s*  did  not  adequately   simulate   the  enhanced  plume  dimensions
 Served during  these tests.

   ISC  moderately  overpredicted  the  observed  concentrations  during
th   kle meteorological conditions and was  in  reasonable agreement with
8^  observed plume widths  during these conditions.   During neutral  and
arj. *e  conditions,  however,  ISC overpredicted  maximum  concentrations
   underpredicted the plume widths.

of  *he finai observation  regarding plume behavior was  the interaction
COQ... .  plume   with   the  forest  canopy.   During stable,  light  wind
».  1tions the  plume  infiltrated  the  canopy  and  the  tracer material
       trapped  within the  canopy.   During  unstable  conditions,  the
     aPpeared to  be  forced over  the canopy.
            real-world  plume  behavior  can be very  complex and  it is
      that  no  dispersion  model  can  adequately  simulate  all  release
  ekal9urations   for all  conditions.   Monitored  data,  such as  tracer
 or  rernents, provide  a  valuable  method of  testing   model  performance
    S^6Cific site3  and  can  be  used  to  guide  model  users in  their
    Cations of  modeling results.
                          PHOSGENE
                           STORAGE
              Figure 1.   Plant and tracer array map.
                                 573
-------
— T"*1
80
60
•
01
w
jE 40
Q
i

ui
£
p
nf 20

0
PLUME WIDTH AND MAXIMA
TEST #1
.

•
• • •




- A A A A A J
A A A
• ' ' • \ } t 1
BO
41
•
a
•a
~*
| 40

UI
X
3

"• 20

0
	 1 	 1 	 1 	 1 	 1 	 1 i
- PLUME WIDTH AND MAXIMA
TEST #2
***** ^ t
A •* •
. X •
*
, •

(




            MAX PRED (A) / MAX OBS (•)
       489  366  380 264 246  409 662  1644 330

Figure  2.   Observed and  predicted
plume width and ratio of maxima for
Test #1.
             MAX PRED (A) / MAX OBS I"    ,g

        S   6   4   23   0   *   '
Figure  3.   Observed and
plume width and ratio of
Test #2.
   60
   60
o
i
   20
       PLUME WIDTH AND MAXIMA
       TEST #3
            MAX PRED (A) / MAX OBS (•)
       78   49   64  65  107  141 201  214 111


Figure  4.   Observed and  predicted
plume width and ratio  of maxima for
Test #3.
                                             80
                                             60
                                          a  20
      PLUME WIDTH AND MAXIMA
      TEST #4
           MAX PRED (A) / MAX
                                                                   OBS <•>
      288  319  307 38S
                     422 »*
                                                                           ,
                                 l"
Figure 5.   Observed and
plume width and ratio of
Test #4.
                                       574
-------
           PROPERTIES OF HOURLY
        RATIONS OF VOLATILE ORGANIC
           AT BATON ROUGE, LOUISIANA
      - Pollack
   as 3
   * Moezzi
!QI ?ms Applications, Inc.
        Valley Road
     ael, CA 94903
      F- Hunt, Jr.

 i,$t t °* Air Quality Planning and Standards
 tesea!lVlronmental Protection Agency
     ch Triangle Park, NC  27711
 'hlSn
 ions  f*r Presents the results of an exploratory data analysis of hourly concentra-
      Vnl3tile organic compounds measured in Baton Rouge, Louisiana over a two-
          Concentrations are extremely variable, ranging from zero to several
      times the median concentration. The upper tail can be approximated by a
 -i,y ^ a .distribution. Relationships between 24-hour average concentrations and
 Srm«Xirnum  1-, 3-, and 8-hour concentrations are examined to determine if 24-
   r»i asurements can be used to estimate peak short-term concentrations.  In
    t^l t'W
     1 ir>ere is too much variation in peak-to-mean ratios for reliable prediction
      'term p^k concentrations from 24-hour average concentrations. A four-way
    Us  Variance model is fit to isolate diurnal, seasonal, and weekly patterns
    L. Or>s of toxic gases from site-specific effects of wind direction. Significant
   *re Vn Concentrations with time of day» day of week» month, and wind direc-
 t  %c   erved» but tnese effects account for little of the observed variability  in
 °xic r^ ?trations.  The results of this study have implications for the design of
       ltoring programs with less frequent sampling.
                                   575
-------
1  Introduction

                                                                          tratio(|i
      The Louisiana Department of Environmental Quality has measured concent'
of 16 species of volatile organic compounds hourly at the state capital in Baton R   j|
since 1984.  These are probably the only such extensive data on hourly concentrate
volatile organic compounds in the United States. The data set provides an opp°rt ^
for interesting analyses of ambient volatile organic compounds concentrations.
paper we present the results of an exploratory data analysis.

      Our analysis of these data is focused largely on the possibilities for making   j$
inferences about hourly concentrations from more limited data.  In Section 2 #e ^
the monitoring methods and discuss the major emission sources that influence tr '   ^
In Section 3 we examine the distribution of the hourly measurements, giving P^f.j, ^
attention to the higher concentrations. In Section 3 we discuss the extent  to fl"1  j,
short-term concentrations can be predicted from 24-hour average concentration • ^
Section 4 we consider how well hourly data can be predicted on the basis of sea
diurnal, and meteorological patterns.  We conclude by discussing potential  futuf
analyses of this unique data set.
                                                                    u
     All of these analyses were performed on all 16 VOCs monitored at the
site. In this paper we restrict our attention to the analyses for benzene; result
pollutants may be found in our study report.
2  Monitoring Site and Data Base Description                                   ,
                                                                         co
     Since 1984 the Louisiana Department of Environmental Quality has bee n
hourly measurements of volatile organic compounds (VOC) in downtown Bato
near the state capitol.  The monitor is based in an office building'and collec*
level measurements. Several major sources of emissions are within a few m1   10
monitoring site. A major state highway circles the downtown area about a fl11 ^
south and east. There are fuel transfer operations on Mississippi River
northwest.  Most important, there is a large industrial complex due north
stretching from about one and one-half mile away to about three miles away-
this complex are two oil refineries and five chemical plants.
     Sixteen volatile organic compounds are measured hourly by dual COItlP,arne
trolled gas chromatographs. Eight hydrocarbon species are measured by a    Oj
zation detector (FID).  They are benzene, toluene, ethyl benzene, m-xylen >   ^
butane, hexane, and pentane. In addition, unknown and total hydrocarbons s'   tje^
sured.  Eight chlorinated hydrocarbon species are measured by a Hall elect
ductivity detector (HECD). They are vinyl chloride, methylene chloride,  1»
ethane, chloroform, ethylene, dichloride, carbon tetrachloride, trichloroet >
perchloroethylene.  Unknown chlorinated hydrocarbons and total chlorinate  • ^ ^
bons are also determined.  Meteorological parameters collected hourly are
wind direction, and ambient temperature.
                                                                     tua
     Although the concentrations are reported as hourly averages, the &c  ^ t
continuous air intake sample collection is 20 mintues for the FID and 25 m
HECD,  The remaining portion of the hour is required for gas chroma
analysis. The  instrumentation is automatically calibrated every day
                                     576
-------
 j
^i   ^ 21 hourly samples per day are actually collected.  The measurement system
    Duality assurance features are described in detail elsewhere.

      °u  tne monitoring site has been in operation since 1984, construction near
          site until September 1985 was thought to result in anomalous and unreli-
       ements; we therefore excluded these observations. In this analysis we exa-
    Ver 11,000 hourly measurements taken between 3 October 1985 and 30 June 1987.
   hi
    Sure 1 shows the relative abundance of each of the hydrocarbon species mea-
        °n avera6e hourly ppbv concentrations. The largest proportion, 33 percent,
        known hydrocarbons. Of the identifiable hydrocarbons, propane and butane
       t for 18 percent of the total.  Next in abundance are toluene and benzene,
          for 13 and 10 percent, respectively, of total hydrocarbons. In this report
     ntrate mainly on the hourly concentrations of benzene, a known human carcino-
1V
   *u

       tions of Hourly Concentrations

      rn°st basic and one of the most useful ways to analyze the approximately
  °t th   measurements °f each species is to try to describe the statistical distribu-
   's th* Concentrations'  This wil1 provide direct answers to such questions as, How
   te  • Dourly concentration of benzene above 10 ppbv?  Furthermore, a statistical
   nc  I2;ation of these concentration distributions might be useful in making
   4    about hourly concentrations at a site where concentrations are not measured


   y jjj ^istribution of benzene concentrations is fairly typical. A cumulative fre-
  'Hrji   uti°n for benzene concentrations below 10 ppbv is presented in Figure 2
    nary statistics are listed in Table I.

         Percent of the measurements are zero: in a large proportion of the hours
       w&s detected at all. Many more measurements are not much  above zero: the
          *S ^'6 pPbv>  the median is  I-8 PPbv' and 90 percent of the measurements
             ppbv.  The highest measurements are very far from zero indeed: the
           PPDV> four measurements are over 500 ppbv, and about 100 are over
          five niSnest values ail occur in January or March, and two of the five
    is  s are on the same January day in 1986.  A cause for unusually high concen-
    arft0"^11 by LADEQ staf f wnen tneX occur.  In most cases, the highest concen-
  ts,  associated with an unexpected release from one of the nearby chemical
K
         ary statistics in Table I indicate that the distribution of benzene concen-
  ..^  aracterized by low levels for the majority of the hours but very high levels
i^tfoH Pr°P°rtion of the hours. Such a heavy-tailed distribution is common for
    J,. ar>t measurements, which can often be well described by a lognormal distri-
      ^re 3 is a lognormal probability plot of the concentrations of benzene. The
       is the van der Waerden normal score, or the value from the standard normal
      ^at would be expected to have a given rank. For example, the 84th percen-
       3 a normal score of one because one standard deviation above the  mean of a
       'ution is the 84th percentile.  The  horizontal axis is the common logarithm
       tration. Values of zero were assigned a logarithm of -1.5.  If the points in a

                                  577
-------
lognormal probability plot fall along a straight line, we can conclude that the rneaS.
ments do indeed arise from a lognormal distribution.  In viewing Figure 3 it should
kept in mind that the single point at abcissa -1.5 represents about 20 percent of tn
and the other Z's in the lower portion of the curve also represent many observation
each. The figure therefore shows mainly the upper tail of the distribution.

     The  points are in a fairly straight line over a considerable range, from benzfi  ^
centrations of less than 1 ppbv (log 1 = 0) to the maximum concentration of 708 pP  ^
708 = 2,85).  Thus the lognormal distribution is a fairly good fit to the empirical dis
tion  in this range.  The bending of the curve at concentrations below 1 ppbv is per   y
not too important.  Although there are a great many concentrations in this range*    \l
are only reported to one significant figure, and may be close to the limit of dete    ^
the measurements were reported to two or three significant digits (as the higher
trations are), then the bending of  the curve at concentrations below 1 ppbv mig"
occur.

     This distributional fit offers the possibility of getting at least a rough idea  ^
behavior of the extreme hourly concentrations from less extensive data. Supp°  ^
only  a few hundred hourly concentrations were available to estimate the 90th a
percentiles.  This would be sufficient to estimate the 90th and  95th percentiles   ^ ^
accurately.  Assuming the logarithms of the concentrations are normal with me  jtf.
standard deviation a, the following simultaneous equations could be solved for v
        y + 1.28o = log PQQ and

        y + 1.65o = log p95 .
                                                                        .
     This amounts to fitting a straight line through the 90th and 95th percenti ®
probability plot. The estimated values of y and a could then be used to est^rna,u
extreme quantiles. For example, the characteristic monthly maximum, the ^.^
exceeded one hour per month in the long run, can be estimated by the upper /
tile of the fitted lognormal distribution. For a standard normal distribution, t  ^
upper quantile is 2.98.  The characteristic monthly maximum could be estirna
antilogarithm of y  + 2.98o after y and a have been estimated.                     ,,
     As a numerical example, consider the hourly concentrations of
in Table I, the observed 90th and 95th percentiles are  10.3 and 20.2 ppbv
0.00 and a = 0.79.  The characteristic monthly maximum would therefore e
log'lO.OO + (2.98-0.79)] = 227 ppbv.  In fact, the observed upper  1/720 quanU
ppbv. The estimated characteristic monthly maximum is very close indeed
dieted by this simple logarithmic fit  to the data. This is not the  case for all  ^ IS  j
VOCs. For example, the actual characteristic monthly maximum for butan >  ^jta
factor of two larger than the estimate of 186. This gives a fair picture of t   tode
of the technique. Obviously this method of estimation is not accurate er l   ge0
mine compliance with a standard.  However, for a distribution spanning a r    ^ 3
centrations from zero to hundreds  of times the median, the method may Pr
estimate of  the magnitude of extreme events not otherwise available.
*  Relationships Between Mean and Peak Concentrations                       ,«
                                                                        r,ttlf
     The relationships between the 24-hour mean and daily maximum °ne" sess ^
hour, and eight-hour average concentrations were investigated in order to
                                        578
-------
**1m
^ easurements of 24-hour concentrations can represent daily short-term peaks.  The
(^  erni maxima were computed using fixed block averaging; for example, eight-hour
   Cations were averaged for midnight to 0700, 0800 to 1500, and 1600 to 2300.
   p
\  °rre^ation coefficients between daily average concentrations and the daily maxi-
studj  e~^°ur, three-hour, and eight-hour average concentrations for each species
    &fa. listed in Table II. The correlations between daily average and peak one-hour
       ions range from a low of 0.53 for trichloroethylene to a high of 0.97 for carbon
    Oride; most correlations are between 0.70 and 0.90.  For benzene, the correlation
          average and peak one-hour concentrations (based on 582 pairs of available
  c     '  Thus, 100(0,85)  = 72 percent of the variance about the mean of peak one-
     Centrations can be explained by the 24-hour average concentration.

  in °ne Would expect, correlations generally increase as averaging periods for the
  ar reases.  Correlations between 24-hour average and peak eight-hour concentra-
N< i, above 0.90 for most species. In general, correlations between 24-hour mean and
      "» and 8-hour concentrations for the hydrocarbons are higher than those for the
     °rganic compounds.
   e
  f  terplots of daily mean benzene concentrations against peak 1-hour, 3-hour, and
      Centrations are shown in Figure 4. The maximum daily benzene concentrations
   Of  erable scatter in the upper fifth percentile of daily means. In the lower
 i'v6lv  data tne ^inear relationship between daily mean and peak concentrations is
 •  fro   r°n*>*  pl°ts *or many of the other species exhibit similar patterns.  As can be
        6 plots» linear regression predictions of peak concentrations based  on the
       Concentrations would substantially overpredict or underpredict many of the
   oUr Values in the high concentration range.  In particular, the two highest observed
      5°ncer>trations, 708 ppbv and 494 ppbv, would be grossly underestimated. Plots
         ~ and eignt-nour benzene concentrations versus daily mean concentration
             than tne one-hour concentrations, as would be expected. Regression
      n       two n^Snest observed three-hour concentrations would be substantially
      j ^ted*  Regression predictions of the highest eight-hour benzene concentrations
      airlV close to the observed values,

       utrerne Points  may strongly influence the correlation coefficient, since they
          -ly to the covariance between the daily mean and peak short-term con-
       '*    Ce» as we observed in the previous section, the distributions of hourly
          5 are verV heavy-tailed and  fit reasonably well to a lognormal distribution,
       lC transformation of the short-term maxima will reduce the influence of the
           ati°nS on the correlation and hence the regression. Scatterplots of log-
      s   tTlean °enzene concentrations against logarithmic daily peak  1-hour con-
   ^6Untare shown in Figure 5. The scatterplot shows much stronger correlations
     Qne.lansformed data*  Regression predictions would still underestimate the
          Ur daiiy maxima, but not as substantially as the regression based on the
           averages.
     ^tiorJ y> correlation coefficients between daily mean and short-term maximum
          " are high; but if a regression line were drawn, the peak concentrations in
           !s of concentration would be highly scattered about the line. Due to this
       SL lnear regression model may often not provide accurate predictions of the
          Ur concentrations from 24-hour concentrations.  Taking logarithms of the
                strengthens the correlations and improves the accuracy of the
         ^iciions, but still may not provide accurate enough predictions of peak
        concentrations.
-------
5  Analysis-of-Variance Models
                                                                            * gad1
     We fit a four-way analysis-of-variance model to the hourly concentrations 01
of the  16 species using SAS PROC GLM to see how much of the variation in VOC c  ^
trations can be explained by simple variables.  The four classification variables we  .jj
of the  day, wind direction, day of the week, and month.  The hours range from zer  e(1t
to 0100) to 23 (2300-2400); three midday hours are usually missing because of instrUaS
calibration and some data are erroneously identified as "hour 81".  Wind direction
converted from degrees into  16 compass points.  A direction of 0 degrees (due nor   ^
DM) was treated as a separate category because calm winds were sometimes reco  ^j
direction 0, and this is different from north winds. There is also an error category  "^
for a few data with direction above 360 degrees.  The days of the week are ere
as 1 to 7, starting from Sunday, and seasonal effects are represented by calendar
month.  We thus have a linear model with 25 + 18 + 7+12 = 62 parameters, whic ^4''
estimated simultaneously by linear least squares. Because the design is unbalanc ^
there are more east winds in some hours than others) the parameter estimates ar   i
simple averages.  The average concentration for a given hour (or wind direction*   ^
the week, month), controlling for the effects of the other variables, is estimate
"least -squares mean" for that hour (or other variable).
     Table III is the analysis-of-variance table for benzene. Each of the four
effects is significant at the 0.01 level, indicating that benzene concentrations   .
differ among hours, days, seasons and wind directions.  Perhaps most remarkaw   j
R  value of 0.09. These effects, though highly significant, account for only a s  ^in-
fraction of the variablity in benzene concentrations. We are therefore left wi     j(
mous variation in hourly concentrations  that cannot be explained by seasonal*
weekly or meteorological effects and appears to be random.
                                                                           Q
     Considering the extent of the range of concentrations, the skewness of t    Ca
tion over this range, and the reasonable  lognormal  fit to the distribution, it se  erltra'
sonable to try fitting the logarithms of the concentrations, rather than the con '  j
tions themselves, as functions of the same four sets of explanatory variables-
equivalent to assuming a multiplicative rather than additive relationship betw
effects of wind speed and hour, for example.  The zero concentrations presen ^ ^
in taking logarithms, of course; we somewhat arbitrarily took the logarithm °
-1.5.
Table IV is the analysis-of-variance table for log benzene. The R  v
                                                                         •
                                                                         j
                                                                  •  lrtrtA.il*   rt
compared with 0.09 for the untransformed benzene concentrations.  This log   ^0!r"
may therefore be a slightly better way to describe the seasonal, diurnal, and  .Q{$ &
logical patterns of concentrations, but most of the variation in  the concentr
remains unexplained.
                                                                         X
     The least squares means for benzene are plotted by hour,  wind direct*  » ^&
and day in Figures 6 through 9, respectively.  These plots show the least sq    .^t ^
resulting from fitting the model to both the transformed and untransforrne
least squares means from the transformed data were back-transformed to   ^
units and are thus geometric means. The geometric means are  always lowe
arithmetic means because they are  much less influenced by the higher cone
                                        580
-------
^  i °ncentrations tend to be highest in the morning peak traffic period, as shown by
*l$o   've*y high least-squares means for hours 6 and 7 in Figure 6.  Concentrations are
\l    iveiy high during hours 20 and 21 just past the evening rush hour.  It is interest-
ti(w n°te that in the hour immediately after the midday calibration  period concentra-
''°ns jte.niSher relative to the following few hours.  This may indicate higher concentra-
   ting the missing midday hours, when presumably traffic increases.

    ^""therly winds are generally associated with relatively high benzene concentra-
  .  North of Baton Rouge is a large complex of refineries and chemical plants. Con-
  ^ tlQns tend to be relatively high in January (three of the five highest recorded
    6 concentrations occurred in January).  Sundays have relatively low concentra-
   as would be expected.

   .  should be emphasized again that these effects, while highly significant, account
     little  of the variation in benzene concentrations. For example, the highest
         squares mean is 9.9 ppbv, and the lowest is 3.6 ppbv.  This would appear to
            difference, practically as well as statistically.  Nevertheless, it does not
          explaining concentrations as high as 500 ppbv.
 ly ]
  - est-s
      ions

      Louisiana data offer significant insights into the patterns of atmospheric con-
      " of volatile organic compounds. However, our exploratory analysis leaves
      Questions unanswered.

      distributions of VOCs at Baton Rouge are characterized by an impressive
    r  '\  A large proportion of the hourly concentrations are recorded as zero; many
    ldatl°ns are in the range from 0.1 to 1.0 ppbv; and several observations are in the
    s of parts per billion. In most cases this  variability can reasonably be compared
 \ Df  °f a lognormal distribution. This comparison can be used  to give a rough idea
      cable maximum hourly concentrations over  the course of a month, for
    lt "°wever, without a better fit or a theoretical reason for the distribution to be
            remains considerable uncertainty about the general applicability of this
V*ti0n     y concentrations for a day are fairly highly correlated with average con-
Cc°ncS f°r the day or part of the day§  This Su88ests tnat measuring, say, 24-hour
l( . Ho    rations wil1 8ive a  fairly good idea of which days have high peak concentra-
 \$ j^ever, it would be very difficult to estimate the  magnitude of the peak concen-
      Ot» the 24-hour means.

V^ilv a°ral and meteorological patterns are  evident in the data.  There are signifi-
      a""' weekly cycles and wind direction effects.  However, these effects account
          " ^action of the variability in the hourly measurements.
^'^ds -Ve ttlus analyzed some of the variation in concentrations of volatile organic
S?* linv? Ways that may be useful for various purposes. Our understanding is at
      M  .   however, and we must conclude that these concentrations have a quite
        ^ined variation. The peak concentrations could be studied in further detail.
             to know whether these individual hourly peaks coincide with identifi-
          >ns of day, time, season, and wind direction.  Statistical classification
                                    581
-------
techniques could be appropriately applied to this problem, rather than the analys'3
variance techniques that were used to study average concentrations.


Acknowledgements

The authors would like to thank Gustave von Bodungen and Jim Hazlett of the I*° .
Department of Environmental Quality for making available the data discussed
report, for providing information on the monitoring site and methods, and for va
assistance in data interpretation.  We also acknowledge the assistance of Bob J°  ^,
our in-house computer expert, for transporting the data from the Louisiana DEQ   .s
puters.  This work was performed under contract to the Monitoring and Data An
Division of the U.S. EPA's Office of Air Quality Planning and Standards.
References

1.   A. K. Pollack, T. J. Permutt and M. Moezzi, "Statistical Properties of
     Concentrations of  Volatile Organic Compounds at Baton Rouge, Louisiana
     Systems Applications, Inc., San Rafael, California, 1988.

2.   LADEQ, "Quality Assurance Project Plan for the Monitoring of Volatile Ot&
     Compounds," Louisiana Department of Environmental Quality, Office 01
     Quality and Nuclear Energy, 1986.

3.   W. R. Ott, "A Physical Explanation of the Lognormality of Pollutant C
     tions," annual meeting of the Air Pollution Control Association, Dallas»
     1988.
                                     582
-------
TABLE I.  Summary statistics for benzene.


No. of hourly observations = 11,306
Mean =5.76 ppbv

Standard deviation = 20.20 ppbv
Minimum = 0.0 ppbv ^f,)

Percentiles (ppbv):

      25th percentile =0.6
      50th percentile = 1.8
      75th percentile = 4.1
      90th percentile = 10.3
      95th percentile = 20.2
      99th percentile = 82.89

Highest observations (ppbv):

      708Jan.  25, 1986
      494Mar.  4, 1986
      462Mar.  6, 1986
      408Jan.  13, 1986
      378Jan.  13, 1986
TABLE II.   Correlations of daily mean and peak 1-, 3-,
and 8-hour concentrations.

       Compound          1 Hour      3 Hours     8 Hours
n-Propane
n-Butane
n-Pentane
n-Hexane
Benzene
Toluene
Ethyl Benzene
^-Xyiene
Unknown HC
Total HC
^nyl Chloride
"e%lene Chloride
iclUoroethylene
poroform
~thylene Dichloride
Carbon Tetrachloride
richloroethylene
erchloroethylene
un known VOC
J*alJ/OC
0.65
0.88
0.86
0.78
0.85
0.88
0.90
0.82
0.83
0.81
0.84
0.83
0.74
0.67
0.85
0.97
0.53
0.82
0.71
0.72
0.80
0.93
0.92
0.86
0.88
0.92
0.93
0.88
0.89
0.89
0.86
0.80
0.63
0.65
0.84
0.94
0.55
0.76
0.76
0.68
0.94
0.97
0.97
0.94
0.96
0.97
0.97
0.94
0.95
0.95
0.94
0.90
0.85
0.93
0.91
0.97
0.87
0.86
0.88
0.88
                          583
-------
                     TABLE III.  SAS analysis of variance results for hourly benzene concentrations.   The four effects fitter!

                     are hour of the day, wind direction (WD_IIYD),  day of the week,  and month.
O1
00
DEPENDENT VARIABLE: 6EN7.ENE BENZENE
SOURCE
MODEL
ERROR
CORRECTED TOTAL
SOURCE
HOUR
UD HTD
DA* ylEK
MONTH
TABLE IV. SAS
effects fitted
DEPENDENT VARIABLE:
SOURCE
HOUEL
ERROR
CORRECTED 10TAL
SOURCE.
rtOU*.
V>M Jilt*.
DF
57
11248
11305
DF
23
17
6
11
SUM OF SQUARES
412176.93801051
4202BB3. 78649806
4615060.72450858
TTPE 1 SS
24255.07590946
62849.34900108
7138.43670781
317934.07639216
MEAN SQUARE f VAt UL
7231.
17435106 19. Jb
373.65609766

F VALUE
2.82
9.89
J.18
77. J5
19
PR > F OF
0.0001 2i
0.0001 I/
0.0040 6
0.0001 11
analysis of variance results for logarithms of hourly benzene
are hour of the day, wind direction (WD HYD) , day of the week,
L5
OF
57
11248
11305
DF
23
b
loglO Benzene
SUH OF SQUARES
1700.25672254
7836.13803157
9538.39475411
TtPt 1 SS
«0.«l«Wtt
^EL


MEAN SQUARE F VALUE
29.
0.

82906531 42.81
69664726
0
F VALUE PR > F DF
13.1
£
,6 0.0001 23
_ \ "1 Q m Q-^O \ fe
^ -1."i O , ft F R
O.OOul 0,
ROOT USE
.33018618
TYPE 111 SS
28579.89394042
32712. 19753839
5896.9297/290
317934.07639216
concentrations .
and month.

PR > F R
0.0001 0
ROOT USE
.83477378
TYPE 111 SS
23S.G&511018
2fc9. 180419*7
SQUARE C.V.
,04*9311 33-J.781U
BENZENE MEAN
5.75678401
F VALUE PR > F
3.33 O.OU01
5.15 0.000!
2.63 0.01SU
77.35 0.0001
The four

-SQUARE C.V.
.17825-1 1944.6175
L5 MEAN
0.04292740
F VALUE PR > F
14.70 0.0001
22.77 0.0001
\o.a& o.ooo\
vi* ,t l o .fe(m\
-------
  Unknown
  hydrocaitx>ns
                                           Propane
                                                    Butane
FIGURE  1.   Hydrocarbon composition by species (based
on average hourly ppbv concentrations).
                                                                      100.0%
                                                                      90.0%
                                                                      0.0%
                                                                                1   23456
                                                                                          Benzene, ppbv
                                        7   8
FIGURE  2.   Cumulative frequency distribution of
benzene concentrations less  than 10 ppbv.
-------
 J.O
1      I
   1 .0 t
s
c
0
r   0.5 »
e      I

      I
-O.i
-1.5
                                                            22:
                                                  II
                                                 21
                                                        122;
                                                       2Z2
                                                     222
                                                    111
                                                   u
                                           222
                                         222
                            2   111
              -1.4
                        -O.B
                                   -0.?
                                              0.4



                                              B«ni«nc
    9i07 065 HIDDIN
                                                         1.0
                                                                                tti
                                                                               t,r
                                                                              Gil

                                                                            LI 10
                                                                         DOJNS
                                                                       CRJN

                                                                     I2YM
                                                                   221
                                                                   1.6
FIGURE 3.   Lognormal probability plot of hourly benzene concentra

("A"  = one  observation,  "B" = two observations;  etc.;  "Z" =  26  or
observations.)
                                          586
-------
 100 !



 100



 Ho








 SlO








 >oo



1*00
                             *
                                A
                                          A                          A
                              A                                   A
                                             A            A
             A             A
                            A   A

 '«0        ,                A                              A   "
          J   A   AA    *       A                  A
          *       MA*
            *  AM* A    A*
   0  s    ........ [[[ ..... ..................... ..... ............

         10  ls  20  25  30   15  40   45  SO   55  60  65  70  75   BO  8S   90  95   100 105  110 115  120
'Oil.
     ..                                Olllj l«»rl)« !>»";»"»
      A
      4>  Daily mean benzene  concentrations  (ppbv)  versus  daily peak one-hour
        tions-   ("A" = one observation;  "B"  = two observations;  etc.; "Z" =  26
       ^servations.)
 to,
 H,
                                                         A


                                                                 A
                A           *  '  A



                 A   A
                       A
                   A      A
             .A     A   A
       ,. A  .*  A A AA A
       »» .  AA  A A A
     *•»
    0    ""•	.                                                         ......
       s  10  ., -*	*				 	
            15   ID  25   10  35   40  45  50   55  60   65  '0   75  60  85   90  45   100  105  110 115  120
           -
        f^     daily mean  benzene concentrations  (ppbv)  versus  daily peak
-------
       12
       ID-
    a.
    o.
    £   6-
    2
01
00      4
00
                                         Arithmetic mean


                                         Geometric mean
         0 I  i  i  »  i ' I '  I  '  i  '  i  '  i  '  ' '  i  '  <  '  '
           0    2   4   6   8   10  12  14  16  18  20  22  24
        Beginning hour


>.  "SstisnatfidL wean, benzene concentration
                                                      10
                                                                      8-
                                                                    t
                                                                      6-
                                                                                              Arithmetic mean


                                                                                              Geometric mean
                                                          DN  N  NNE NE ENE E  ESE SE  SSE  S SSW SW WSW W WNW NW NNW  N
                                                                                                  Wind Direction


                                                                      FIGURE 1.   Estimated irean benzene concentration (ppbv)  by
                                                                      •wind dxrection.
-------
O1
00
CO
                                                  Arithmetic mean

                                                 Geometric mean

              Jan  Feb Mar Apr May Jun Jul  Aug Sep Oct Nov Dec

                                 Month


        FIGURE 8.  Estimated mean benzene concentration
         (ppbv)  by month.
                                                                        &   5
8
o>   4
CD
                                                                           3-
                                                                           2-
                                                                           1-
                                   Arithmetic mean

                                   Geometric mean
        Mon    Tues   Wed   Thurs    Fri     Sat     Sun

                          Day of Week


 FIGURE 9.   Estimated  mean benzene concentration
 (ppbv)  by day of week.
-------
EVALUATION OF HIGH-VOLUME SAMPLING TECHNIQUES
FOR THE DETERMINATION OF CDD/CDF IN AMBIENT AIR
     C. Tashiro(l), R. Clement(1), A. Szakolcai(2),  W. Chan(2)
(1)  Laboratory Services Branch, Ontario Ministry of Environment
     125 Resources Road, Rexdale, Ontario, Canada, M9W 5L1
(2)  Air Resources Branch, Ontario Ministry of Environment
     4th Floor, 880 Bay St., Toronto, Ontario Canada, M5S 1Z8

     Recently, interest in the analysis of chlorinated
dibenzo-p-dioxins (CDDs) and chlorinated dibenzofurans (CDFs) in
ambient air has increased substantially.  Some measurements of
2,3,7,8-TCDD in ambient air using high volume sampling with 8la3^/< 2)
filters and polyurethane foam (PUF) cartridges have been re ported t  '
Currently, validated methodologies for the sampling and analysis Oj
full range of CDD/CDF congeners in ambient air have not been repo^
Proper method validation requires spiking studies to determine
breakthrough of both the filter and the PUF cartridge and recovery
studies to determine extraction efficiencies.

     The Ontario Ministry of Environment ambient air sampler for
CDDs/CDFs is a modified high volume sampler that utilizes a filt
followed by a single or dual PUF cartridges.  The sampling effic
of the device has been determined by performing low and high ~L&i
CDD/CDF surrogate spiking experiments.  l3C-labelled standards we
spiked separately onto the filters and PUF-cartridges to determin
breakthrough.
                                                               ale
     Results from a series of experiments using filters and sing
dual PUFs indicated good surrogate recovery from both the filter
the PUF.  There was some breakthrough of the lower chlorinated
congeners from the filter to the PUF, but there was no observed
breakthrough of the surrogates from the first PUF to the second
when dual PUFs were used.

Introduction

     The importance of the analysis of dioxins and furans in ^
air has increased over the past few years.  These compounds h   Q
detected in air, not only in urban areas where there are known
such as incinerators, but also in remote locations far removed
                                   590
-------
1 PR?^ source.  To analyze these types of samples, detection limits below
of ^   are required.  To obtain these low detection limits, a great deal
 t    ^as t)een done on both tne analyttcal methodology for sample
   °tion and cleanup and the sample collection methodology.

   r>    e are a number of sampling technologies available.   The samplers
   y°rate a particulate filter and an adsorbent material such as silica
      ~2 resin or PUF to trap vapour phase molecules.   Smith originally
     Slass fibre filter followed by silica gel in a removable
           Silica gel is easily handled and cleaned but is restricted to
           volumes and is subject to humidity related problems.  Smith
     y rePorted the change to PUF as his adsorbent of choice(l).  PUF is
         most widely used adsorbents in ambient air monitoring of
         le organics because it is inexpensive and easy to handle.
       CDD levels have been monitored by the U.S. Environmental
       n Agency using a PUF plug in a low volume Model PS-1 sampler(2).

       °itario Ministry of the Environment has been evaluating a high
     Sampling system for the analysis of total dioxin and total furan
     **3 in arabient atr-  A modified high volume sampler with a
     °°ated glass fibre filter and a single or dual PUFs has been tested
     ^ne collection efficiencies and congener breakthroughs.  Spiking
       ts were carried out at various levels and for various sampling
    01'  Anotner objective of the Ministry was to evaluate the analytical
        y for fche determination of CDD/CDF in filter and adsorbent
          is led to the dsv610?™6^ of sampling and analytical protocols
   o=     in embient air.  A comparison with a Model PS-1  sampler was
    arried out.

           Methods

       Pup material was purchased from a local upholstery manufacturer
   Pu a density of 24.03 Kg/m3.  The teflon-coated glass fibre filters
   ^lafia3ed from Pallflex Products Corp. (Putnam, Conn.),  All solvent
   a p^lled-in-glass grade (Caledon Laboratories, Georgetown, Ontario).
     eii  and alurnina were obtained from BioRad Labs (Richmond, CA).
           dioxin and furan standards were obtained from Cambridge
                   (Cambridge,  MA.) and Wellington Laboratories (Guelph,


    fll?lfied hign volume sampler (Anderson shelter) with a single glass
    aiM   and 3in8le or dual PUF cartridges and a Model PS-1  sampler
 Vin8s    al Works> Cleves,  Ohio) were used.  The filter and cartridge
 vMo  Were assembled and spiked in the lab and then taken to an urban
               3ite*  A range of 13C-labelled dioxin and furan standards
          °n tlle filters and cartridges at varying levels.  Samples  were
        °Ver< varying time periods and temperature and flow rate data
    by n!!!0nltored.  All housing, filters and PUF cartridges were proven
             prior to use.
               were refrigerated after collection.   The filters and PUFs
           extracted for 24  hours with toluene.   An internal  recovery
           spiked  prior to extraction.   After extraction,  the samples
   '^sij4,1^ up using a modified Dow cleanup with  NaOH/silica,
        ""•  AgN03/silica and alumina adsorbents.
                                 591
-------
      The GC/MS  analysis was  carried out on a Finnigan 4500 GC/MS/DS
 (Sunnyvale, CA.) with a 30m  DB-=5 column (J+W Scientific, Inc., F
 CA.)  in the selected ion monitoring mode.  Percentage recoveries
 breakthrough of the congeners from the filters and from the cart
 were  calculated.

 Results

      An initial set of spiking experiments was carried out in the
 of  1987.  Four sets of ambient air samples were collected using a mo1
 high  volume sampler.  The sampling conditions are listed below:

      Sample A:  Filter + Single PUF + High Level Spike (24 hours)
      Sample B:  Filter + Single PUF + Low  Level Spike (24 hours)
      Sample C:  Filter + Dual   PUF + High Level Spike (72 hours)
      Sample D:  Filter + Dual   PUF + Low  Level Spike (72 hours)

 The 13C12-labelled congeners that were used and where they were sj
 indicated in Table I.  13C,a-6CDD was used as an extraction effici.f0n of
 spike.  There was good recovery for most congeners with the except. $
 13C12-8CDF.  At low spike levels the 8CDF spike was not recovered    j
 high  spike levels, recovery was greater than 200?.  This high reCT0
 have  arisen from the use of only one extraction efficiency spike J
 correct all field spike recoveries.  The filter spike recoveries
 I reflect the total spike recovery, i.e. the total amount of SP
recovered from the PUF and filter that was originally spiked °n..,^ef  i
filter.  Table II indicates the percentage breakthrough of the f*   \&
spikes onto the PUF.  There was a high degree of breakthrough of  ^
chlorinated congeners (tetra and penta) from the filter to the P^ oCtar
little breakthrough of the higher chlorinated congeners (hepta 3°
There was no observed breakthrough of any of the *3C12-labelled
from the filter or the first PUF to the second PUF when dual *"
used.
     A number of native dioxins and furans were also detected i
spiking experiment samples.  These samples were collected in an^
Ontario location.  The range of positives detected and detectio
are shown in Table III.  All congeners were detected in the
although not in the same samples at the same time.  A lar^er-.und
higher chlorinated congeners were detected, with 8CDD being " °j
eleven of twelve samples.  8CDD is generally found in most env
samples.  The median range of detection limits for the PUFs ^
0.01 - 0.3 pg/m3 for all congeners.
                                                                 pf&
     A comparison of MOE's modified high volume sampler and a   ^ t
Metal Works Model PS-1 sampler was carried out in December 1?   g of
same urban location.  Table IV compares the percentage ^ecove
13C12-labelled congeners spiked on both samplers.  In this e^eS
four analytical recovery spikes were used.  The total recove ^
that there were no major losses from the PUFs or the ^ *lterSonto
samplers.  There was some breakthrough of the filter spikes  ,^0
in both samplers.  There was more breakthrough of the lower
more volatile congeners from the filters in both samplers.

Conclusions                                                        A cfl
                                                                   A
     There was good overall recovery of the labelled
both the high volume and PS-1 samplers, however some

                                   592
-------
field  spike recoveries  was  observed.   The high volume sampler has the
advantage of being  able to  sample a larger volume of air in the same time
period as a PS-1  sampler and  thus achieve lower detection limits..  There
was  breakthrough  of lower chlorinated congeners from the filter to the PUF
in the spike range  investigated.   More breakthrough was observed in the
summer, when the  spiking experiment was carried out (Table II), as
compared to the winter,  when  the  sampler comparison took place (Table IV).
Although the sampling protocol  has not been finalized,  it will involve the
use  of a single PUF and filter  in a modified high volume sampler with a
21-1(8  hour sampling period.   The  analytical protocol used was determined
to be  efficient for these types of samples as reflected by analytical
spike  recoveries  in the range of  50-  '\HQ%.

References

1.   R.M. Smith,  P.W. O'Keefe,  D.R. Hilker,  K.M.  Aldous,  S.H.  Mo,
R.M. Stelle, "Ambient air and incinerator testing for chlorinated
dibenzofurans and dioxins by  low  resolution mass  spectrometry",  presented
at The Seventh International Symposium on Chlorinated Dioxins and  Related
Compounds.  Las Vegas, Nevada, 129.  (1987).

2.   B.J. Fairless,  D.I. Bates,  J.  Hudson,  R.D. Kleopfer,  T.T.  Holloway,
D.A.  Morey,  T.  Babb, "Procedures  used  to  measure  the amount  of
2,3,7,7-tetrachlorodibenzo-p-dioxin in  ambient  air  near  a superfund  site
cleanup operation",  Environ^ Sci.   Teohnol.  21_:  550. (198?).

3.   R.M. Smith,  P.W. O'Keefe, D.R. Hilker,  K.M.  Aldous,  "Determination of
picogram per cubic meter concentrations of  tetra- and pentachlorinated
dibenzofurans  and dibenzo-p-dioxins in  indoor air by high-resolution  gas
chromatography/high-resolution mass spectrometry",  Anal.  Chem. 58:2414.
                                  593
-------
Table I.  Spiking Experiment - Average Percent Recoveries*
 Congener
 I 3
  C12-4CDD
 13C12-5CDD
 13C12-7CDD
 1 3
  !C12-8CDD

 13C12-4CDF
 13C12-5CDF
 13C12-7CDF
 13C12-8CDF
IIO(P)
170(F)
170(F)
160(P)

 93(F)
 97(P)
150(P)
260(F)
                                     B_

                                   92(F)
                                  160(P)
                                   77(P)
                                   7KP)
                                  190(F)
                                   ND(P)
170(F)
97(P)
120(P)
130(F)
130(P)
130(F)
77(F)
260(P)
74(F)
INT(P)
170(F)
170(F)
120(P)
1 1 0(F)
72(F)
ND(P)
NOTE: A = HIGH SPIKE (5-1Ong) - SINGLE PUF - 24 HR
      B = LOW  SPIKE (.3-.5ng)- SINGLE PUF - 24 HR
      C = LOW  SPIKE (5-1Ong) - DUAL   PUF - 72 HR
      D = LOW  SPIKE (.3-.5ng)- DUAL   PUF - 72 HR

      P = SPIKED ON PUF
      F = SPIKED ON FILTER (RECOVERED FROM PUF AND FILTER)
     ND = NOT DETECTED
    INT = INTERFERENCE
     *  = RECOVERIES CORRECTED FOR SINGLE ANALYTICAL RECOVERY SPlKE
Table II.  Spiking Experiment - % Breakthrough of Filter Spikes Onto
Congener Group       ^

TETRA               89
PENTA              110
HEPTA               10
OCTA                ND
 B

61
36
12
ND
                                               170
                                               100
                                                11
                                                ND
                                           68
                                           69
                                           ND
                                           ND
NOTE: A, B.C.D - SEE TABLE I
      ND - NOT DETECTED
      *  = RECOVERIES CORRECTED FOR SINGLE ANALYTICAL RECOVERY
                                   594
-------
 ble Hi.   Spiking Experiment  -  Native CDD/CDF*

SSfigENER
RANGE DETECTED   MEDIAN
                                     #  POSITIVES/
DETECTION LIMITS
5CDD
?CDp
(PG/M3)
0.1-0.2
0.08
0.01-2.9
0.02-7
0.05-47
0.04-6
0.04-0.7
0.1-2.8
0.01-3.9
0.09-3.9
(PG/M3)
0.2
	
0.05
1.3
1 .5
0.6
	
0.5
2.0
0.6
# SAMPLES
3/12
1/12
3/12
6/12
11/12
7/12
2/12
6/12
5/12
6/12
(PG/M3)
0.009-0.2
0.009-1.7
0.04-1.5
0.06-0.9
0.09-1.9
0.02-0.2
0.007-0.2
0.007-0.3
0.007-0.9
0.007-1.3
         RECOVERIES CORRECTED  FOR  SINGLE  ANALYTICAL  RECOVERY SPIKE
         Sampler Comparison - % Recoveries**

          WHERE  SPIKES   TOTAL %  RECOVERED

                          HIVOL    PS-1
                                            % BREAKTHROUGH

                                        HIVOL         PS-1
FILTER
FILTER
PUF
PUF
PUF
130
78
99
160
88
79
99
100
140
150
5(5/6)
3(1/6)
	
	

11(5/6)
6(4/6)

	
____

           , # BREAKTHROUGHS (X) PER NUMBER OF SAMPLES  (N)
          FOUR ANALYTICAL RECOVERY SPIKES USED TO CORRECT DATA
                                 595
-------
EVALUATION OF THE COLLECTION EFFICIENCY
OF A HIGH VOLUME SAMPLER  FITTED
WITH AN ORGANIC SAMPLING  MODULE FOR COLLECTION
OF SPECIFIC POLYHALOGENATED DIBENZODIOXIN
AND DIBENZOFURAN ISOMERS  PRESENT IN AMBIENT AIR
T. 0. Tiernan
D. J. Vagel,  G.  F.  VanNess,
J. H. Garrett,  J.  G.  Solch
and L. A. Harden
Wright State  University
175 Brehm Laboratory
Dayton, OH 45435


     Laboratory  tests with a  high volume ambient air  sampler,  which
incorporates  a glass fiber filter and a  cartridge  containing  polyurethane
foam (PUF) plugs and XAD-2  resin,  have shown  that  this  sampler  effectively
traps and retains  78-100% of  each of the  2,3,7,8-chlorine-substituted
dibenzo-p-dioxin (CDD)  and dibenzofuran (CDF)  isomers,  when these  compounds
are applied as pure compounds  to the filter,  prior  to  sampling 250-300 m3
of air. However,  some  of  these compounds migrate from the filter to the PUF
cartridge in  the source of  sampling.  The same behavior  is  observed when a
synthetic flyash spiked with  these isomers is applied to the filter prior
to air sampling.  However,  when  these isomers are applied to the  filter in
the form of natural incinerator  flyash containing  the  compounds,  migration
of the CDD/CDF from the filter  to  the backup  cartridge  does not occur to a
significant  extent during air  sampling.  Following  evaluation of  the
collection efficiency  of  the sampler, ambient  air samples were  collected in
a metropolitan area in which MSW incinerators and  other  potential airborne
sources  of CDD/CDF are  located.   These  samples  indicated that  airborne
CDD/CDF  congeners  are  present  in  the  ambient  air  in  the industrialized
metropolitan  area at concentrations in  the range from 0.03 to  12.7 pg/m3.
Patterns  of  CDD/CDF isomers observed  at  some  sampling sites  in  these
studies  compared  well  with those  reported to  be  observed in  the  stack
effluents from  municipal incinerators  sampled  elsewhere.   No detectable
concentrations  of CDD/CDF  were  found   in samples collected  at  rural
locations in  this  study.
Introduction

     Previous experiments conducted  in our  laboratory  have  demonstrated
that a high volume  ambient air sampler employing a filter,  two  polyurethane
foam  (PUF)  plugs  and a solid sorbent effectively retains each  of the
2,3,7,8-substituted  polychlorinated  dibenzo-p-dioxin  (CDD)  and
                                   596
-------
     or mated  dibenzofuran  (CDF) isomers.1   In  these experiments,
       COD and CDF isomers,  applied  as  pure compounds to the  filter  of
   »orx a*r samP,ler»  *ere  observed to migrate   from  the  filter to  the
of grii beQ.t cartridge in the course  of  collecting air  samples.   The  extent
*°fe hfav*on was 0reater  *°r the lower chlorinated  species  than for  the
CDf ^Oaly chlorinated  CDD  and CDF  isomers.   It  particulate-bound CDD  and
^let r frs are also stripped  from  particulates  which are trapped on  the
*hiC]j nllter °* such an air  sampler, then  the interpretation of  compounds
t«pre P**8  the filter  and  are  collected  in the  PUF/sorbent  section  as
'fr°&      ' those  present in the vapor phase in the  air sampled  would  be
     ?c^nt rat ions  of  CDD and CDF  in  ambient  air samples have also  been
       *& a typical metropolitan area  using the  sampler  described  above.
           t"°  large municipal solid waste (MSW)  incinerators,  as well  as
       otner potential sources of air-borne CDD and CDF.   The ambient air
      .Ct>Hected  in  this study were intended to  provide  preliminary
       °n On CDD/CDF levels present in background ambient air,  as well  as
            in air in industrially-impacted areas.
         	    The ambient  air  sampler  used for these experiments  was
         Metal Works, Inc. Model PS-1 PUF sampler.  The sampling module,
            a QM-A quartz filter (Whatman Laboratory Products,  Inc.)  and
         cartridge, has been described elsewhere.1'2
             and  Precipitator  Flyash.    A synthetic flyash was  prepared
                   of 38% aluminum  oxide, 39%  iron (III) oxide and  23%
           oxide. The  synthetic  flyash was  prepared, thoroughly  mixed,
           the  2,3,7,8-substituted  isomers,  dried overnight  and  mixed
         two hours.   A  natural  flyash  sample,  collected from the
         "c  precipitator  of  a  municipal  incinerator,  was obtained  from
         Ministry of the Environment,  Toronto,  Ontario,  Canada.  The two
          (synthetic and real) were analyzed prior to use to determine
           CDF concentrations.   In two separate  sets  of experiments,
         of the  synthetic ash and  the  MSW  incinerator f lyashes were
    the»  to tne  inlet  filters of  the sampler prior  to air sampling.
   *** »*» San>Ples were in operation, other samplers containing  unspiked
           operated to monitor the  ambient  air  for  background  CDD  and
                     and Collection Procedures.    Air  sampling  studies
       a ncil&erator  and laboratory-prepared  synthetic flyash were
Si*1 »it? Ascribed  for a previous  study.1  Sampling was conducted  at
o ji* coij*8 »aich are  used by the regional air monitoring agency.   During
\ * bour*;ction, the sampler was operated at the maximum flow  rate for  22
SlS*Plin periods- for  transportation to  and  from  the  sampling locations,
        n       -
      ift K tttodules were  wrapped in aluminum  foil,  the cartridges were
 X j* in Btle.s "ith aluminum-foil  lined  caps and  the filters were
    tt«a»ti   tri Dishes sealed with teflon  tape.   The collected  samples
         p?rted to  the  laboratory  in  ice  chests  maintained  at  a
               20" C.
                     .  Isolation and  GC-MS  Analyses  Subsamples  of  the
       for  *8h and the  precipitator  flyash were  Soxhlet extracted  and
           CVD and CDF.   The filters and cartridges from the  laboratory
            3encV  tests  were separately  extracted  (again using  Soxhlet
    *     ,
    a Coy  .an«  analyzed,  in order to  determine  the   distribution of  the
     *bUht80tt.ers between the two trapping media.  For  each  of  the actual
             .
         iH air  samples,  the  filter and cartridge were removed from  the
           »ers' and placed  together  in  a  single Soxhlet extraction
           ni 8ollition  containing  one 13Ci 2 -labeled isomer of  each
        «w ***  of CDD  and  CDF was  added  to  the  contents of  each
      '   ?L  e  samples  were extracted for  a  period of  16  hours with
              sample preparation,  liquid  chromatographic  clean-up
            GC-Ms analysis procedures have been described previously.1-3
                                 597
-------
Results

     Synthetic Flyash Experiment.   To investigate the effect  of saj?£.j| <*'
synthetic  flyash-bound  CDD and CDF,  the  spiked  synthetic £ 1**voW
distributed on the inlet filter of  the  ambient air sampler.   A toW* vCf»F
of 334 m3 of  air  was drawn through the sampler at 230 L/min.  ™e-»t W
temperature during  sampling was 14°C  and the  relative  humidity  » ^ fiJ'
Table  I  shows the percentages of the  2,3,7,8-substituted CDD- ijoifij
isomers  remaining  on the  filter  and retained  in the cartridge *•  „ .W
correction for the background  ambient  air levels.  These results "j,^*!*
same preferential  migration of the  lower chlorinated isomers as*nt«r.S
earlier when the 2,3,7,8-substituted isomers were applied to  the i  flD IP
pure compounds.1   For each  isomer,  however,  the  percent  retain*«i).  J?t
filter is higher  when the  isomers  are present  on the synthetic  * ^e*"*,
previous experiments involved the  collection of 271 m3 of air at » flrat*'|
higher  average  temperature  than  here  (26°C) , and  this te"-.^oB ?,
difference may  account  in part for the observed  trend in Biffr  ^ t*
isomers. The  total  recoveries  of  the  isomers  are  in  agreement    flj t»
previous results which demonstrated highly efficient retenti°i
2,3,7,8-substituted CDD and CDF isomers by the ambient air sampler*
     MSW  Incinerator  Flyash Experiments.   Another  experiment  »l   Oj
that just described was designed to test the effects  of  air  samp^iJutfl}w<
CDD and CDF  components  of  natural flyash.   The flyash was distr   $ u
the inlet filter of the sampler and 346 m3  of air were  collected  t9llfy
rate  of  238  L/min.    The  average  temperature  was  8°C and  tn 0yeS^rt
humidity was 36%.  The results for the MSW incinerator flyash  are v  $ #e
in Table II.  Only very small quantities of the CDD  and CDF conffjw ^
observed to migrate from the  spiked filter  to the cartridge.   TD  ttat«Jj
of these experiments suggest that the  CDD and CDF  isomers sorbea  ,^<;is ^
flyash behave differently  from  CDD and CDF isomers  which are **   ^iJ*.^
introduced into a synthetic inorganic matrix.  Further  experiioen gaJ0pJj*f(
required  to investigate  the effects of  flyash  composition/  ^p*
conditions and particle size on the distribution of  isomers  in cn      jft

     Ambient Air Samples.   Eight ambient air samples  were collect ^
of an initial survey of the metropolitan area and surrounding c
The sample numbers, collection sites and sampling conditions  **"*   _
in  Table  III.   The  concentrations  of  CDD and CDF for the  YJ^t* JV
collected at the rural  site were  below or near  the  detection £* ^/»  d
averaged 0.04 pg/ra3 for tetra-, penta- and hexa-CDD/CDF and °*?ieCt«° X
hepta- and octa-CDD/CDF.   Ambient  air  samples 5 and 7  were cOAattd »  i*
the intersection  of  two heavily  traveled  streets.  Samples *    af**^
collected within  1  km of  a MSW incinerator.  The  downtO1>^j,re«'2
represented by samples 2 and 4 which were collected  on top of  &va^\e J  (
building.  The data for samples 2, 4,  5 and 6 are  presented  in
                                                               .
     Several interesting comparisons can be made from the data 1
Figure 1  indicates that the total CDD  and  CDF congener
distributions which are characteristic of air samples collecteo  jfejj|*'
of the MSW incinerator and in the   suburban-roadside  area  r**f.f-ato* f
quite  different.   A careful examination of  the GC-MS  c°
however, shows a similar CDD and CDF isomer pattern for  each        -
congener profile for  the  suburban-roadside site does suggest   ^e c v .
source of OCDD may  be influencing  this site.  A comparison or   f"
CDF total congener  profiles for  samples  2 and 4 is presented * bij
Sample 2 was collected following several days of clear weather  ra
4 was collected at  the  same  site on the following day durl,n?   tbe
which  continued during most  of the  sampling  period.
concentrations are different on the  two days,  the p
similar for these two samples collected at the same site.

     A comparison  of the congener  profiles  displayed in
shows that the samples  collected in the downtown area are v e"ar**;fl« J
those samples collected in the   more immediate  MSW Incinerator att*'
is also apparent in the GC-MS chromatograms where the isomer *
the samples  from each location  are nearly  identical.   Table

                                   598
                                                          F^rV **   *$
                                                           v e"
-------
        fxf !ihe 9DD.and CDF  congener  profiles obtained  in the  present
        *be Msw incinerator  area and the  downtown  area with previously
   DP i-  sults for MSW  incinerator  related samples.  The patterns of CDD
 r»t tn°»ers  in the  air samples collected in the present study are very
 !l «a th     e observed  stack  emissions  from a MSW incinerator stacks, as
  11 a H*  incinerator  in  another  study.
*
 VK
*ft Ji,.2'3'7'8~substituted CDD and CDF  isomers  are added to the  filter
!:'*aic fiSam£     as pure  compounds,  or  incorporated into  a  synthetic
   to fi     '  significant migration  of  the  added  isomers  from  the inlet
   ked  »*«.v  /S0rb8nt   car.tridsre  is  observed.   However, when the  filter
   n»i*r,.rj-   natural MSW  incinerator  flyash containing  CDD/CDF,  very
   iff Cation from the  filter to the backup cartridge occurs during air
      Attbient  air  samples collected  in the metropolitan  area  showed
      *e variability in the  concentrations  of the congener groups, the
      nepta-CDD/CDF having the highest concentrations,  while  levels of
      "t  in ambient  air  from the  rural  area were essentially  non-
      •  Air samples  collected in the  metropolitan  area  have a  common
       n*r!  but  samP1inff sites  within  the study  area  appear  to be
       CDD and CDF from  more  than one source.
     Da*
    dlSr n.5
             , T.  0.  Tiernan,  M.  L.  Taylor,  J.  H.  Garrett,  G.  F.
              G.  Solch and L. A.  Harden,  "Assessments of  ambient  air
          techniques for collecting  airborne  polyhalogenated dibenzo-
          s  (PCDD),  dibenzofurans   (PCDF),   and biphenyls (PCB)  "
            ,  (1988), in press.
            and  M.  D.  Jackson,  "Modification  and evaluation  of  a
           air  sampler for  pesticides  and  semivolatile  industrial
       chemicals," Anal. Chem. 54: 592 (1982).

             G.  L.  Ferguson, T.  0.  Tiernan, G.  F.  VanNess,  J.  H.
            1. Wagel  and M.  L.  Taylor,  Chlorinated Dioxins and
              in  the Total  Environment  if,Butterworth  Publishers,
            ,  pp.  377-397.
          D*
                              Table  I
             of  2,3,7,8-substituted isomers  following  air  sampling
             spiked  synthetic  ash  is placed on  the  filter.
Percent
Recovered
in Filter
7
31
59
75
99
75
85
4
21
20
55
63
80
90
85
110
69
Percent
Recovered
in Cartridge
77
53
8
7
0
0
0
91
75
53
32
33
3
0
0
0
0
Total
Percent
Recovered
84
84
67
82
99
75
85
95
96
73
87
96
83
90
85
110
69
                                 599
-------
                                 Table  II
    Distribution of CDD  and  CDF  congener  groups  following  air  sampli0*
            when MSW  precipitator  ash  is  placed  on  the  filter.
       Congener
       Group
                      Percent
                      Remaining
                      on Filter
                   Percent
                   Migration
                   to Cartridge
       TCDD
       PeCDD
       HxCDD
       HpCDD
       OCDD

       TCDF
       PeCDF
       HxCDF
       HpCDF
       OCDF
    Locations
                        100.0
                         99.7
                        100.0
                         99.4
                         93.6
                        99.
                        99.
                        99.
                       100.0
                       100.0
   ,9
   .3
   .9
        0.0
        0.3
        0.0
        0.6
        6.4

        0.1
        0.7
        0.1
        0.0
        0.0
                            Table  III
            and  sampling  conditions   for   ambient
                              air saa:
                     pie*'
wsu
Sample
Number
1
3
2
4
5
7
6
8
Site
(a)
7
7
1
1
13
13
8
8
Date
(1988)
4/5-6
4/6-7
4/5-6
4/6-7
4/7-8
4/8-9
4/7-8
4/8-9
Volume
(m3)
358
378
398
390
360
375
384
377
Flowrate
(L/min)
254
270
276
287
272
280
285
290
%RH
34
82
34
82
66
52
66
52
Average
TeoP
t^cl
22
8
22
8
7
8
7
8
(a)   Site  7 - Rural area, 15 km NE of metropolitan area
      Site  1 - Downtown metropolitan area
      Site 13 - Suburban Roadside, 8 km SE of metropolitan area
      Site  8 - MSW Incinerator Area, 6 km SW of metropolitan area
                               Table V
      Comparison of CDD and CDF congener ratios to published
 Congener
 Class
          Percentages of Total CDD and CDF in Each Congener Class
                     Present Study             Published Results
                               MSW         Incinerator
CDD Congeners
  TCDD
  PeCDD
  HxCDD
  RpCDD
  OCDD

CDF Congeners
  TCDF
  PeCDF
  HxCDF
  HpCDF
  OCDF
             Downtown
              Site I
Incinerator
  Site 8
  Stack
Emissions*
                 3
                 7
                13
                38
                39
                 6
                20
                38
                36
                ND
      1
      7
     26
     38
     28
      3
     14
     36
     36
     11
     3
    10
    17
    28
    42
    15
    24
    31
    25
     5
a
b
Hay, et. al., Chemosphere: 15, 1201  (1986)
Rappe, et. al., Cbenospbere: 16, 1975  (1987)

                               600
Incin
      1
     12
     26
     26
     36
     33
     3*

     8
      $
-------
0}
o
                                 Table IV    Concentrations of PCDD/PCDF in ambient
                                   air samples (concentrations in pg/cubic meters)
                              Congener
                              Group
Sample 6
Suburban
Site 13
                                          (0.11)
Sample 6
MSW Area
Site 6
Sample 2
Downtown
Site 1
                                                                    (1.30)
Sample  4
Downtown
Site 1
TCDD

PeCDD
HxCDD
HpCDD
OCDD
TCDF
PeCDF
HxCDF
HpCDF
OCOF
0.03

ND
(0.04)
0.14
0.47
a ee
NO
(0.02)
0.12
0.61
NO
(0.07)
ND
0.26

2.18
7.60
11.07
8.31
1.23
5.10
12.66
12.71
3.78
0.44

1.09
2.07
e.ii
6.18
1.48
6.00
9.70
8.17
ND
ND
S0.06)
ND
(0.07)
2.03
2.59
4.96
0.30
2.19
4.18
4.06
ND
                                      (0.29)
                                                 TCDD PeCDD HxCDD HpCDD OCDD TCDF PeCDF HxCDF HpCDF OCDF

                                                                       Figure 1
                                                        Comparison of Congener  Profiles  from
                                                           Sites 8 and 13 for the  Same Day


                                                    cone, in ps/cubic meter
                                                                                          TCDD PeCDD HxCDDHpCDO OCDD TCDF PeCDF HxCDF HpCDF OCDF
                                                                                                                Figure 2
                                                                                                 Comparison of Congener Profiles from
                                                                                                 Downtown Area on Two Different Days
-------
 MONITORING AMBIENT AIR FOR DIOXINS
 Billy J.  Fairless,  Ph.D. and Jody L. Hudson
 Environmental Monitoring and Conpliance Branch
 Environmental Services Division
 U.S.  Environmental  Protection Agency
 Kansas City, Kansas
     Procedures are described for designing a monitoring network,
collecting representative sanples, and then corrparing the neasured
concentrations to the applicable criteria.  Results from the collection
and analyses of over 1,000 sanples are used as a basis for certain
conclusions and for reconmendations to improve cost effectiveness when
monitoring for 2,3,7,8-TCDD in ambient air.
                                  602
-------
          VII has had a need to monitor for dioxins in the environment
        early 1970's when the 2,3,7,8-congener was first found in
         In 1985 we started the removal of contaminated soil from the
       these sites and, in the process, our cleanup activities became
        sources of 2,3,7,8-TCDD in the ambient air.  A strategy to
      for the amount of 2,3,7,8-TCDD in the air was developed and irnple-
     Qt that time.   Since then, air monitoring has been conducted at
   ^•Qrge sites where the cleanup extended over tine periods of a few
    to several months.  We have obtained additional information from
    studies and formed some new conclusions about the process.

   .^ objectives of each of these monitoring activities were to measure
       mtration of 2,3,7,8-TCDD at the site boundary and to compare the
       concentrations in the ambient air with the applicable criteria.
      steps were taken to meet these objectives.  The first step was to
        the applicable criteria and then agree on a set of procedures
        ng measured concentrations to the criteria.  Once these items
   CQnfcleted,  the design of a monitoring network proceeded.

       S.  Margolis,  Center for Disease Control (CDC), provided us with
    estimated no observed effect level (NOEL) of 5.5 pg/MJ for exposure
           of a few nonths.  This number was adopted as the applicable
        and the lower concentration of 3 pg/M3 was also adopted as a
      level.
       next step in the Process was to establish procedures for relating
       Concentra t ions to the criteria.  One of the first decisions to be
      this phase was to decide how concentrations below the method
        limit were to be treated.  We elected to treat these
   to  as roeasured concentrations (see reference 1) .  A second decision
       tablish a time period for averaging measured concentrations when
       nSi a mean exposure concentration to be compared to the criteria.
         competing considerations that were balanced when making this
         Since the applicable criteria was based on an assumed health
             from a lifetime exposure, it would be desirable to make the
        Period as long as possible.  However, it was also desirable to
         report any high concentrations back to the field crew as
      as Possible so timely pollution abatement actions could be taken.
            were discussed with CDC staff during the time they were
       g their recommendations for a ICEL, and are reflected in the "few
t  ^exii?lrase of their recommendation.  The criteria allows sane degree
«S bn?llity for the monitoring agency to set a practical averaging
 ^n    at the sams tirre»  it contains the appropriate time constraining
     1  Initially,  we elected to use a 14-calendar day averaging
        We encountered several problems with this decision and are now
           that,  when possible, longer averaging times be used.
                   Design of the Monitoring Network
      r.
        tunatelv'  a11  site Perimeters do not have a symmetry approxi-
       ac^uare or circle.   Figures 1,  2 and 3 show three sites having
          nt  shaPes and requiring different designs for the monitoring
             monitoring network and rationale for site 1 was provided
           During that cleanup,  all  monitors were operated at all

                                 603
-------
 times  cleanup activities were occurring.   For the cleanup at site 2, *
 initiated the practice  of  only operating  that part of  the air monitor
                                                                    ® '
 network that was in the vicinity of  the cleanup operations.   For
 these  operations occurred to the east  and west  of Rock Creek Road  (see
 Figure 2) .  Four monitors were operated during  cleanup on the east si
 the  road.  Four monitors were also operated during cleanup of the
 side of the road.   Site 3 was particularly difficult because the
 nation was primarily along the sides of roads in an urban area.
 required that  the air monitors be placed very close to the cleanup
 tions or  that multiple access agreements be obtained.  We  selected a
 of  12 monitoring locations  for  this  site which, as with all other si
 were not  moved during the cleanup operation.  The cleanup  activities
 passed very close  to many of these locations during cleanup of the si ^
 As  with site 2,  site 3 was  divided into multiple sections.  Initial1^
 monitors  were operated.  The monitors covering a particular section
 site were turned off as soon as a section was determined to be clean*

     The  primary concern when establishing the number of monitoring
 locations in the network was to insure that a plume of contamina^ .
 did not pass undetected between the  monitors.  Frequent changes in
 direction, longer  sampling  times, greater numbers of sampling I02
 lower method detection limits and cleanup over large areas within a  j(
 all tend  to decrease the probability that pollution will be uec
 Since starting to  monitor for dioxins in ambient air, we have
 longer sampling  times (now  72 hours) which also result in  lower
 detection limits (now 0.8 pg/Mr).  These steps increase the
 that any  pollution will be  detected  by the monitoring network.

     The  decision  to use 14 data values from each monitor  to ca^°T^
 mean ambient air concentration  is based on the probability that t^e
 measured  mean concentration will not be significantly different ("
 fidence level, data with a  relative  standard deviation of  24% ar
 distribution) from the true value1.  We continue to believe that
value are needed to obtain a reasonably reliable estimate  of the
 ambient air concentration of 2,3,7,8-TCDD.  Initially, we
 200 samples for  both particulate materials and 2,3,7,8-TCDD.  We
unable to establish any relationship between the two variables.
 fore, we  no longer analyze  samples for particulate matter.

Collection of Representative Samples

     Samples are collected on filters and polyurethane foam usin£re6
General Metal Works, Inc., Model PS-1 sampler and operating
described in reference 1.  Approximately 1,200 cubic meters of
 sampled during the 72 hours we now use to collect each sample-
Comparability

                                                               nti^
     The mean and the 95% confidence limits of the 14 most recent
generated concentration values from each monitor are compared to ^
criteria and warning values of 3.0 pg/M , respectively  (Figure   *  j_g
calculating the mean concentration, the measurement detection I1
used as a measured concentration for all concentrations that are
the detection limit.
                                   604
-------
         Procedures
       filter and polyurethane are ccnibined and extracted together in a
      apparatus.  The extract is cleaned with silica gel, alumina and
  c    -and then ana^yze<^ using capillary gas chroma tography and low
13 , AJJtion or tandem mass spectroscopy .  Quantitation is based on carbon-
  Aabeled internal standards.

      Assurance

2,3 7A generic quality assurance  (QA) project plan for monitoring for
It   l8jrCLD in ambient air has been written and approved by Region VII.
    now used for any site where air monitoring for 2,3,7,8-TCDD is
      f and is always supplemented by a site-specific sampling plan.
       specific sampling plan provides any unique objectives and covers
       n °^ the sanpiinQ network for each of the specific sites.  Taken
        he two documents address each of the data quality variables
        in EPA guidance for preparing QA project plans3.
          Practical description of our general procedures for monitoring
          g the quality of the data we generate may be found in the most
    Oi   f t °I ItGuidance Document for Assessment of RCRA Environmental
    H^litv''  or in the computer software user's guideb we use to calcu-
    ^ Quantitative data quality indicators.

       Sutle concentration of 2,3,7,8-TCDD was observed, the air
      *<*s very close (within 100 feet) to the location where earth was
     ^r?? Or V)here tord materials (concrete, heavy equipment, steel,
      *e ha^ decontaminated using a stream of high-pressure water.

                                 605
-------
 Five additional sanples from site 2 had detectable concentrations of
 2,3,7,8-TCDD.  Like site 1, each of these sanples were collected
 sanpler was both very close and downwind of equipment cleaning
 Only two sanples collected at site 3 had detectable concentrations
 though the nonitors were frequently very close to the cleanup opera
 From these results, we conclude that there was not a measurable bac
 of 2,3,7,8-TCDD in the general area where these sites are located.
 We also believe the procedures are sufficiently sensitive to detect
 quantify concentrations at the NOEL of 5.5 p/M3.
      Levels of 2,3,7,8-TCDD in soil at the three sites where air
 toring was performed were similar having concentrations in the range
 approximately 1 to 600 ug/kg.  Although the currently available data
 insufficient to establish a correlation between concentrations of
 2,3,7,8-TCDD in soil and background concentrations in air, the da^J
 show that measurable concentrations in air will probably not be found
 soil concentrations are in this range.  However, before broader con
 sions can be drawn regarding background concentrations in air at all
 sites (approximately 35) awaiting cleanup, sate of which have wides]
 concentrations of 2,3,7,8-TCDD in soil in excess of 1,000 ug/kg,
 additional data will be necessary.

      We conclude that two kinds of cleanup activities will cause the
 ambient air to become polluted with 2,3,7,8-TCDD if they are not *#
 properly.  The most serious of these activities involves the use of
 or high-pressure water to clean hard surfaces.  From site observatic*03'
 appears that an aerosol is formed by such operations which contains
 contaminated materials that remain suspended long enough to reach
 sanpler.   Corrective actions are simple and,  in our experience,
 They are to orient the hard surface being cleaned so the stream of
 is reflected down instead of up or to enclose the spraying area.
 activities that generate dust are less serious,  but also need to be
 controlled.  An effective control is to insure that the soil  is
 using fire hoses or lawn-type sprinklers as necessary.   Although
 concept,  this control is not easily carried out during the hot
 months when large quantities of dirt are being taken at different
 from the surface.
     On only one  occasion, during the three  years we have been «•—-
for 2,3,7,8-TCDD, did the running average  exceed the 5.5  pg/M3 NOEL-
exceedance was caused by one  sample collected at site 2 on  Se
1986, which  raised  the average  to between  7  and  8 pg/M .  The
tion resulted from  the use of high-pressure  water.   Corrective
were taken iinmediately upon notification of  the  single high cou—
thus avoiding an  extended or  elevated exceedance.  The area in the
ity of  the monitor  was unpopulated so an exposure at the  NOEL for
of a few months did not  occur.  The concentration averages  calculi
the remaining samples were all  below  the action  level of  3.0 pg/M
these results, we conclude that the probability  of a cleanup opera
a site where  soil concentrations  do not exceed 100 ug/kg, causing a
fleant amount of 2,3,7,8-TCDD to  be in the ambient air, is very
now believe that a  greatly reduced network is  adequate to detect
leased air pollution.  For future cleanup  operations where  2,3,"
concentrations in soil do not greatly exceed 100 ug/kg, we will is" ^
that one or two monitors, maintained  in a  predominately downwind P°L4j
relative to each cleanup activity, will be sufficient unless  and V&
measurable concentration is obtained.


                                   606
-------
   summary,  the method described above is quite rugged.  It is a
     and cost-effective way to monitor for 2,3,7,8-TCDD in ambient
    when used properly, will give data of unusually high quality at
     low concentrations.

 ne mention of trade names is for informational purposes only and
    Constitute an endorsement or recommendation by EPA or the
1*  B.  j.  Fairless,  et.al.,  "Procedures to measure the amount of
    2,3,7,8-tetrachlorodibenzo-p-dioxin in the ambient air near a
    superfund site cleanup operation, " Environmental  Science and
    Technology,  Volume 21, No.  6,  page 550 (1987).

 •  J.  L.  Hudson,  "Quality assurance project plan for air monitoring
    at  dioxin remediation sites, "  U.S. EPA,  Kansas City,  Kansas,
    (March 10,  1987).

3*  U.S. EPA,  Office of Research and Development,  QAMS-005/80.

 *  B.  J.  Fairless and D.  I.  Bates,  "Guidance document for assessment
    of  environmental data," U.S. EPA,  Kansas City,  Kansas (February
    1988) .

 *  B-  J.  Fairless and M. A.  Ibmpkins,  "Labor and sample  tracking
    (LAST)  computer  software,"  U.S.  EPA, Kansas City,  Kansas
    (February 1988).

 *  J.  L.  Hudson and D.  A. Morey,  "Evaluation of method performance
    for measuring  2,3,7,8-tetrachlorodibenzo-p-dioxin  in  ambient
    air;"  Proceedings  of Dioxin 87,  Seventh International  Symposium;
    Chemosphere, in  press.
                             607
-------
O)
§
                                                                          	 PROPERTY LINE

                                                                           	 WATER LEVEL
                                                                          	 ROM!
-------
50
     100
              200 FT.
                               Figure 2   Site 2 - Each side of this
                               site contains approximately one acre.
•*  WATER  LEVEL
—  ROAD
-  FENCE
    SAMPLER  LOCATION
-------
                                      Di ox i n
                                 Cleanup
Air Monitor Lo
                                                cal'0"'
Figure 3   Site 3 - The
boundaries of this site
generally parallel the
major roads in the area.
                       610
-------
120 -i
          Mean Percent Recoveries
                      versus
       Increasing  Sampling  Duration
                                          A 2.3.7.6-TCDF
                                          x 1.2.3.4-TCDD
                                          O 2.3,7.8-TCDD
       Duration In Hours(Volume In cubic meters)
                                         72
                                        (1080)
                   Figure S
-------
        14 DATA POINT  RUNNING  AVERAGE
w • «* w
3.00
o^ 2.50
O It.
oo
L^.
^^
1 0=
ODul
rJE 2.00
•a
-o
2 ~=
NJ oo 1.50
CrtO
•<*<
OZ 4 MA
o^ 1.00
•— W1
a.
0.50
0.00
mm
mm

^M
tm
mm

mm
mmm
mm-
i^Hi

•-
•

m/

m


m

























<











i
,













































<











<
i











i

1
	 WAMHLQJ.EVH,







,.
**WttWlltttt|{||ff||{|:

— 1 — 1 1 1
••








::


mm










I
mm








<


mm








M


mmm








i


^mmmmm







II
1;

_l
^








!<


IM








l(


^








K


^








i.


^mt








(


^^•.•^^••^








9XS9XX

1 	 L
0 	 ,»
x Mtniu
. •M-SAWLIM










,            ,
                  cX. tt«a \4 roost TO=en.tiy ooUecbeA
                                                    axA
-------

     q    °N °F POLYCHLORINATED DIBENZO-p-
         D DIBENZQFURANS IN STACK GAS
         AND AMBIENT AIR
   irt
 Jeth0t}  *  Harleas and Robert G. Lewis
  vli-o   evel°Pment Branch
                  oring Systems Laboratory
                  Protection Agency
                 Park,  North Carolina 27711
                and
    -on    Bldg*  1105
    En^ntal Chemistry  Laboratory
    BfeT ronmental Protection  Agency
        l°n, Mississippi  39529
 NUg
'faction   deacribes  the quality assurance /quality control  procedures,
*L*& i«e   and cleanup procedures  and high resolution gas  chromatography
TU  ^aat   '10n mSS sPectrometry (HRGC-HRMS) method of  analysis used for
*<}  ns (PCDpn\°f  ^o;i^chlorinated  dibenzo-£-dioxins  (PCDDs) and  dibenzo-
!;  fOr f       in several sample  matrices.   MM5 sample trains were operat-
   '  Th    hours to collect  samples  of  stack gas emissions  during  each
    gage  lri
-------
 Introduction

       Polych.lorinated dibenzo-p_-dioxins  (PCDDs) and dibenzofurans  (^ Of
 enter  the  environment  by  two routes,  manufacture, use  and  dispos0-1 •  .
 specific chemical products and by -products;  and  from the emissions
 specific types  of  combustion and  incineration  processes.   Atrnosp"'  e
 transport  is  considered a major  route  for widespread dispersal of   .g
 compounds  in  stack gas  emissions into  the environment.  Various   ®
or concentrations of PCDDs and PCDFs nay be found in stack gases,
scrubber water emissions  from combustion processes.  The PCDDs and
are found as  complex  mixtures of many congeners and  isomers and us1   g
                                                                   3
in low concentrations.  Those tetra-, penta-, and hexa-CDD and CDF i3  ,e
with  chlorine  in the 2378 positions are  considered to be the roost * fll
isoraers, but  usually account  for  only a  small percentage  of "the   ^
concentrations  of respective congeners.  In our studies the 2378-TCD » ^
of 22 TCDD isomers  and the most toxic  member of the PCDDs and PCDFs'ijue
usually a  very  minor component  of respective total concentrations-  ^
123^-TCDD  and many other isomers and congeners are  considered to be   ftj
tively non-toxic.  Therefore, extremely sensitive and  specific ana-w  .
procedures are  required  for isolation, identification and quant if
of PCDDs and PCDFs in order to provide the quality of data that is
for meaningful assessment purposes.  For example,  (l) total concentfa  8
of respective  PCDD  and PCDF  congeners and  (2)  concentrations f°r   it)
specific tetra-, penta-,  and hexa-CDD  and  CDF isoraers  with chl  ^ol"
the 2378 positions,  isomer specific analysis.  These are the most
tant isomers because they  are used  for determination of toxic   u*
values that  are used  for  health risk  assessment  purposes.   These
type isoraers  are retained  in  tissue of  lifeforms such  as  human8'
and wildlife.

                                                                   W ^
      A Congressionally -funded National Dioxin Study was  initiated W ^e
U.S. Environmental  Protection  Agency  (EPA)  in   1983  to  determin ^
magnitude of  2378-TCDD and other PCDD  and  PCDF  contamination in ^gtl0"
sample matrices.  Tier k of this study addressed  emissions from c°^ e3s^'
and incineration processes.  Stack testing was performed  on 13 Pr° ,J$>'
These were sewage sludge  incinerators, kraft paper  recovery boil61*'
trial waste  incinerator,  wire reclamation  incinerator,  secondary
smelter, carbon regeneration furnace, drum and barrel  furnace, wo°  g
wood-fired boiler, and irobile sources.  Ash samples from 75 pr*ocesS
also analyzed in this study (1,2).  Extracts of ash samples from 8?cn
combustion processes  were  also  subjected  to  analysis  for br°   a0
dioxins and dibenzofurans  (3)«  The analytical methods and quali^ ^
ance/quality control procedures used by  theEPA's Environmental M°n  [
Systems Laboratory,  Research  Triangle  Park (EMSL-RTP)  and
Chemistry Laboratory,  Bay  St. Louis  (ECL-BSL)  in these  studie3
scribed elsewhere (k). The analytical methods were first used in

                                                               A  &$* If
      Attention has been  focused on  determination of  PCDDs  ana  ^fgff /
ambient air  only in recent years  (6,7,8).  These  analyses are e  -$
difficult because they must be performed at the  concentration •*•  \i^
femtogram to low picogram-per-cubic meter (pg/rn^) of air.  Minii"11
of detection  of 0.05 to O.U pg/m3 mist be achieved in  order t°
data that has meaningful  and significant value  for assessment
The two EPA laboratories are now participating in a one-year C°nSjgf)t *>
ally-funded study  for determination  of PCDDs  and PCDFs  in  a^^g Ju
around a  new and  highly  efficient municipal  incinerator that
been put into operation.


                                   614
-------
ECI(   Ttle analytical  methods and  QA/QC procedures  used by  EMSL-RTP and
4v*i1    *n  the  anibient  air  study are described  elsewhere (9)  and are
V0fv  ^e upon  request.   The study is  briefly  summarized.   General Metal
f^ 3, s~!  samplers equipped with quartz-fiber  filters  and polyurethane
lg   |pUF)  are  used for  collection of samples.  The  quartz-fiber filter
PS^P  ed with  °-8  ng    Ci2 123^-TCDD prior to operation to determine the
51er C°:L:Lec'tion anii retention efficiency  for  PCDDs and PCDFs.  The  sam-
\ * are operated  for 2k  hours  and collect 350 to 1*50 m3 of ambient air.
Of g llter  and  PUF  of each sampler are combined, spiked with  ng  amounts
fid V]?arton  labeled PCDD and  PCDF  internal standards and Soxhlet extract-
thg     Benzene for 16 hours.   Clean-up  of extract is accomplished using
c^pi c*d/base  procedure,  micro  alumina  column  and  followed by  a micro
e*tr°n Column-   A  2.5  ng amount  of  3'Cli|  2378-TCDD is  spiked  to the
to g«   derived from  the  carbon column and prior to  final concentration
Of i,  ^  for analysis.    This standard is used to  determine the  amounts
*»„. ® 8 carbon  labeled  internal  standards in  the extract,  the method
   ^lency.  High   resolution  gas  chroraatography-high  resolution  mass
       bry  (HRGC-HRMS)  analyses are  performed using a Varian/MAT  311A
Itl j|7'1 efflploya  the  multiple ion  monitoring technique.   The  MS is operated
*Vjl  EI  lnode*  T^e  GC  is equipped with a 30 m  SE-5U or  60 m  SP-2331
for a  33-^ica  capillary columns.   An  SS-200 PDF  11/32*  data  system  is  used
     ^uisition and processing of  data.

^lUi    etluate  safety and  safe  handling procedures  in the  laboratory are
    red when working with 2378-TCDD  and  other PCDDs and PCDFs.

     ^e carbon-labeled  PCDD and  PCDF  standards  are  used as internal
        to  determine  method  efficiency  and  for  identification  and
 "Hv   ltion  purposes  using respective response  factors  developed  with
^54 f standards,  such as those shown.  The carbon-labeled  123^-TCDD is
     0 determine PS-1 sampler efficiency.

      Labeled                               Native
,r.  .,   —.                               2378-TCDF
1?1  -1    TCDD                               2378-TCDD
                                            12378-, 23l*78-penta-CDF
                                            12378-penta-CDD
          -hexa-CDF                         123678-, hexa-CDF
  ^"^3678-hexa-CDD                         1231*78-, 123678-, 123789-hexa-
n »
  l9"123lt678-hepta-CDD                      123l*678-hepta-CDD
 .-,  ^                                      1231*678-, 123l*689-hepta-CDFs
 ctf^DD                                   OCDD and OCDF
           standards of  specific isomers  or all  tetra- through  octa-
       CDp3  are used for identification purposes.   Examples are  shown
       « 1 through 1*.
              » cleanup and analyses are performed on a "set" of samples
      rig of:
                                615
-------
- 9 or 10 test samples
- Field blank
- Method blank
- Laboratory fortified sample or control sample.
                                                                    j.£V6
A quantification standard containing exact amounts  of labeled and n  eS
PCDD and PCDF standards used in  preparation of test samples and QA t**®
is prepared with each "set" of samples.

      Analytical criteria used  for  identification  and  confirms-*10
2378-TCDD and other PCDDS and PCDFs include:

- Correct HRGC-HRMS  retention time  (+.  3 sec) of 2378-TCDD
  labeled 2378-TCDD on  an  isomer specific column.  This also a
  the labeled and  native  2378-TCDF,  penta-,  hexa-,  hepta- and
  and CDFs.

- Correct chlorine isotope ratio of molecular ions (_+ 20$)

- Correct HRGC-HRMS multiple ion monitoring  responses for exact
  of PCDDs and PCDFs

- Responses  of  molecular ions  mist  be  greater  than  2.5  x noise
                                                                   dc
- HRGC-HRMS retention time  windows of respective series  of CDD a
  isomers
                                                                     t
- Comparison  of sample  analysis  with  analysis  of standard con
  all or specific PCDDs and PCDFs.
-Supplemental  criteria such as  COC1  loss,  determination of  e  rlfl
  compositions of molecular ions in real time  and analysis  to  coni  e
  absence of specific chlorinated  diphenylethers are performed,
                                                             -
Examples of criteria used  in  analysis  is  evident  in  Figures -t
                                                                   l
      The QA/QC  criteria and requirements  for analytical  data

      Criteria                               Requirements

o  Method efficiency achieved for:
   carbon-labeled tetra- through hexa-
   CDDs and CDFs                             50 to 120$

   Carbon-labeled hepta- and octa-
   CDD                                       UO to 120$

o  Analytical criteria used for              Satisty  criteria
   confirration of PCDDs and PCDFs           previously stated

o  Accuracy and precision achieved           50 to 150$
   for determination of PCDDs and
   PCDFs in laboratory fortified samples,
   control samples, duplicate of test
   sample*
                                  616
-------
                                           Re quirement s

    hod blank and control matrix           Described below
  ^ee of PCDD and PCDF contamination
    target mininum limits of detection
  The +
     target  minimum  limits of detection  for PCDDs and  PCDFs  that are
                                                      Stack     Ambient
                                           Ash       Emission'8-' Air^)
                                          (ppt)       (ng)      (pg/m3)

^ Tcn°DD and 2378-TCDF                        1        0.2         0.1
1U f1 isomers and 37 TCDF isomers           10        0.3         0.2
1H6   CDDs and 28 Penta-CDFs               20        0.3         0.2
^h^^008 and 16 hexa-CDFs                 20        0.1*         0.3
           and ^ hepta-CDFs                 UO        0.5         0.3
       OCDF                                 50        0.5         o.U

    M5 sample train operated for 1* hours.
    s~l sampler operated for 2k hours, 1*00 ra? air.
      e results for PCDDs  and PCDFs  in ash and stack gas emissions are
    ri
     2e(i in Table I.  In-depth evaluation of results for Tier k ash and
            provided  elsewhere  (l,  2).   The highest concentrations and
       Ppm and  micrograras , of tetra- through octa-CDDs and  CDFs were
  e ,   in ash and  stack gas from secondary  copper  smelters.   With the
        °f W00d stove and in°feile sources, PCDDs and PCDFs were detected
  8ta v
  to v  8as  emissions  from the other processes  in  amounts ranging from
 bQg hiSh  ng.   Our  procedures are  not adequate  for determination  of
    and  PCDFs  in emissions  from  wood  stoves  and  mobile  sources.
    T*!"i
      e analytical procedures are performing  well and achieving sub-
   .    7nuw limits of detection  for PCDDs and PCDFs  in the ambient air
   *  Ampler locations,  wind direction,  many data points,  and a tho-
      sil-uation  of all analytical and  meteorological  data will  be  re-
   ct  n  Orcier to  isolate the  source or  sources of PCDDs  and PCDFs
       in the ambient air study.
I,
     BaUonal Dioxin Study," U.S. Environmental Protection Agency,
     ffice of Solid Waste and Emergency Response, Vfeshington, DC  20li6o,
    Deport No. EPA/530-SW-87-025, August 198?.
s
              Dioxin Study, Tier h Combustion Sources Project Summary
          ," U.S. Environmental Protection Agency , Office of Air Quality
             and  Standards,  Research Triangle  Park, NC 27711,  Report
      «  EPA-U50/l*-8U-Ollt,  September 1987.
                                617
-------
      R. L. Harless, R. G. Lewis, D. D. McDaniel and A. E. Dupuy,    ^
      "Identification of Bromo/Chloro Dibenzo-p_-dioxins and Dibenzof^r Qf)
      in Ash Samples," Proceedings  of Seventh International
      Chlorinated Dioxins and Related Compounds, Las Vegas, NV. In P
      Chemosphere , 1988.

      "National Dioxin Study, Analytical Procedures and Quality Assurft
      Plan for the Analysis  of Tetra through Octa Chlorinated Dibenz  X^
      dioxins and Dibenzofurans in  Samples  from  Tier h  Combusti°n
      Incineration Processes,"   U.S.  Environmental  Protection  Ag  ^
      Environmental Research Laboratory,  Duluth, MN   558UO.  Addend
      Report NO. EPA/600/3-85/019, May 1986.

      R. L. Harless, A. E. Dupuy and D. D. McDaniel, from "Human and ^«
      Environmental Risks of Chlorinated  Dioxins and Related Comp°un
      pp. 65-72, Plenum Publishing Corporation (1983).
6.    R. A. Kites, B. D. McVeety and J. M. Czuzwa, Science, pp.
      November (198*0 •
                                                                   62T
7.    K. Olie, M. Berg and 0. Hutzinger, Chemosphere, Vol. 12, pp*
      (1983).

8.    M. Oehne, S. Man, A. Mikalsen and P. Kirschmer, Chemosphere,
      15, pp. 607 (1986).

9.    R. L. Harless, R. G. Lewis, D. McDaniel and A. Dupuy, "Anaiy*1^ Ot
      Procedures and  Quality  Assurance Plan  for  the  Deterrainatio^l
      PCDDs and PCDFs in Ambient Air  Near the Rutland,  Vermont ^^^
      Incinerator," U. S. Environmental Protection Agency, Environ  ^
      Monitoring Systems Laboratory,  Research Triangle Park,  NC
      Unpublished report, April (1988).

Disclaimer
                                                                  reflflCt
      The research described in this article does not necessarily   ^d*
the views of  the  Agency and no official endorsement  should  be in  .$$*
Mention of  trade  names  or  commercial  products  does  not  cons
endorsement or recommendation for use.
                                   618
-------
    Il  Summary of Analytical Results  for Tier It, National Dioxin Study.
               Compounds

               23T8-TCDD
               Other PCDDs  and  PCDFs

               2378-TCDD
               Other PCDDs  and  PCDFs
                                       Concentrations/Amounts
                                          Ranges  Detected

                                          ppt to  ppb
                                          ppt to  ppra

                                          pg to ng
                                          pg to y g
   378-TCDF, penta- and hexa-CDD and CDF isoraers with chlorine in 2378
  * sitions usually minor comonents (l to 30^) of respective  total
  Cor*entrations.
     1234689-
        A
    429.774
1234678-
                       '    423.777
12MWSr A1234679-  •*«•   *••
       JLf 1234689- _^T2347B9-
      Z"'r''**?7*"'"'""*'"	
      88 !•••• it'M t»>M l»-39 2**M 20
                                                ttC1fOCDD
                                                              471.773
                                                 OCOO
!**»'   49f.739


       497.738
                                                   PFKuf.
                                                       MBS*    443.74*
                                                 OCOf
                                                          *    441.743
s.
            The analysis of hepta CDDs
            and CDFs on a 30m SE-54
            column.
             "Ctt-2378-TCDD.
                                 I
                                       22'W 24'94'M 3C'M  M>M 4O
                                    Hgina.   The analysis of OCDD and
                                           OCDF on a 30m SE-54 column.
                                                   MASS'    333.934
TCODs
                 JLJL
       2378-
       A
                                                   MASS'    321.894
                  JLJL
                                                   MftSS*    319.697
                                     '•••
                                   13C,r2378-TCDF
                                                            317.939
    5?      ,     luild
 .^^
                                uuu
                                                   MASS'  A  303.902
                                                   30*ee  32 -ee  3*
    ^w* 3. An txpandid vtow of ttw TCDFt and TCDDi ihown In figur* 4
                               619
-------
                                  WINDOW 1
                                        13C,,-12378-ponta-CDD
                                         .,                MASS

                                         wCir1237B-penta-CDF —*
                                                                   	
                                           26-ee  ze>ee  38>ee  32-
                                  WINDOW 2
isoe-
teee-
 see-
                                                                      rf
   Rgur* 4. HRGC-HRMS •natyslt of • standard containing an tetra through htxa PCDD» »nfl

          on a 60m SP-2331 fused silica column. Used for identification purpous.
                                       620
-------
    OMPARISON STUDY of
   AM*,A|R DIOXIN/FURAN SAMPLING
   ANALYTICAL METHODS
r-o
                .Lao
       Canada
      er Road, Ottawa, Ontario

        - Clement, A. Szakolcai and W. Chan
'No 'nistrVof Environment
    •Ontario
       a referer|ce method development project for ambient air dioxin/furan
      p  ' a ^'e'c' samP'm9 mtercomparison study was carried  out jointly by
|£°ntQ *!Jt Canada and the Ontario  Ministry  of  the Environment (OME) in
|l& VPS i   6 samP|er tvPes (two custom modified  hi-vols and a General Metal
P$,s fibre J.sarnpler) were operated in duplicate.  All samplers used teflon coated
V Sar^Dl     S and a Polyuretnane foam (PUF) backup sorbent. Samples from the
      Qth and one moc^'^ed hi-vol were analyzed by the Ontario lab and samples
     H  er modified hi-vol were analyzed by the Environment Canada lab. Both
       essentially  the  same analytical  protocol.  Results from the  inter-
      a  Wi" 'De Presentecl in terms of the overall variability between two
       lencies using a hi-vol/PUF sampling method for dioxin/furan at the same
          Council of Resource and Environment  Ministers (CCREM) is a joint
             Steenn9 committee set up to develop a co-ordinated response to
                  needs in Canada.  One of the  activities of the Council  has
    t0^?nsorsnip of joint federal/provincial research and development activities
\v' ^nd oth measurement.  assessment and control of toxic contaminants in air,
\ih*°-p-H    media.  Because of the intense public interest in polychlormated
Se ^ds [ >X'ns W'ox'ns) and polychlorinated dibenzofurans (furans), this class of
  ects    nas received the highest priority related to research and development
                                621
-------
Development of standardized sampling and analytical approaches for dioxins and
furans is obviously necessary prior to the collection of monitoring data. Beginning
in 1987 the provinces of Alberta, British Columbia, Ontario and Quebec in  co-
operation with Environment Canada  began a program to  develop a reference
method for the measurement of dioxins and furans in ambient air.  Prior to 1987
both Environment Canada (EC) and the Ontario Ministry of the Environment (OME)
had begun work on ambient air monitoring systems for dioxins and furans which
employed high volume samplers  with polyurethane foam (PUF) adsorbent traps.
The analytical  laboratories  of each  organization also had over 10  years of
experience .in the analysis of dioxins  and furans in various  matrices (work with
ambient air samples began in 1986). Based on the ongoing EC/OME work and the
published work of others1-2-3 the proposed reference method was to be based on hi-
vol/PUF sampling and gas chromatography (GC) - low resolution mass spectrometry
(LRMS) techniques. In December of 1987 a field study was carried out to compare
ambient air results obtained  using the dioxin and furan sampling and analytical
methodologies of the Ontario Ministry  of Environment and Environment Canada.

Experimental

                    Design of Field Intercomparison Study

The objectives of the intercomparison study were (1) to determine the variability of
dioxin/furan  ambient air concentrations measured by two different  agencies
employing essentially the same sampling and analytical methodologies (hi-vol/PUF
sampling  and  GC-LRMS analysis); (2)  to determine inter  and intra-agency
measurement  precision, and (3) to determine the adequacy of  quality
assurance/quality control practices.

Two each of three different sampler designs were operated in the field comparison
study.  The samplers consisted of an OME custom modified hi-vol, an Environment
Canada custom modified hi-vol and a commercially available  sampler, the General
Metal Works PS1. The samplers were operated over a 24 h sampling period on three
different days. Additionally a field blank (passive exposure of filter and PUF for 24
h) was collected before and after the active sampling period. Because of limitations
in the analytical budget for this project, additional active sampling could not be
carried out.  Samples from the PS1 and OME modified hi-vol  were analyzed by the
OME, Lab Services and samples from the EC modified hi-vol were analyzed by EC,
Analytical Services  Filter and  PUF samples were analyzed separately for each active
and passive sampling day.

                               Sampling Site

The sampling was carried out near the OME laboratories in the north west section of
metropolitan Toronto  (Rexdale).  All samplers were installed at ground level and
were spaced approximately 5 m apart.  A 200,000 vehicle per day expressway was
located 100 m north of the monitoring location.

                                Sampling

As shown  in Table I the major differences between samplers  were flow rate, filtei
media type and the depth of the PUF sorbent bed.  All samplers used teflon coated
glass fibre filter media.  The Pallflex TX40H120WW filter media used by  EC
                                  622
-------
has a higher collection efficiency than the T60A20 filter used by the OME." The PS1
samplers operated at the lowest average flow rate (240 L/min.} and the EC modified
hi-vol at the highest (850 L/min).  Samples were collected between Dec. 14 and Dec.
23, 1987. Mean temperatures on the active sampling days ranged from 2°Cto -5°C.

 TABLE I -CHARACTERISTICS  of SAMPLERS used in  INTERCOMPARISON STUDY

Type
Filter
Adsorbent
Flow Rate (L/min.)
Sample Volume
(m3) (24 h)
Flow Measurement
Device
1 - EC
Custom Modified
Hi-Vol
PallflexTX40Hi20WW
Teflon Coated Glass
20x25 cm
Polyurethane Foam
(PUF)
Firmness Factor: 31
Density: 24.0 kg/m3
Size: 15 cm x 7.5cm D
-600
800 - 900
Dry Gas Meter
(Temp. Compensated)
2 - OME
Custom Modified
Hi-Vol
PallflexT60A20
Teflon Coated Glass
20x25 cm
Polyurethane Foam
(PUF)
Firmness Factor: 30
Density: 24.0kg/m3
Size: 7 5cm x 8.6cm D
425-600
625-825
Rotameter
3 - OME
General Metal Works
PS1
Pallf!exT60A20
Teflon Coated Glass
10cm Dia.
Polyurethane Foam
(PUF)
Firmness Factor: 30
Density: 24.0 kg/m3
Size: 7.5 cm x 5.9cm D
-280
320-400
Flow Venturi
Magnehelic Gauge
                             Sample Handling

The pre-cleaned PUF sorbent loaded in  canister assemblies and the filters were
installed just before the start-up of the  samplers.  Upon completion of the 24 h
sampling period, filters were removed from the samplers, wrapped in pre-cleaned
foil and sealed in plastic bags. Canister assemblies were capped in the field; PUF
plugs were removed from the canisters in the laboratory, wrapped in foil, sealed in
plastic bags and frozen until analysis.

                                Analytical

The laboratories  employed  nearly identical analytical techniques.  Each discrete
sample (filter or PUF) was spiked with a mixture of isotopically-tabelled surrogates
(2-6 ng per compound) just prior to extraction. The EC lab used tetra through octa
dioxin surrogates while the OME lab used  tetra and penta-furan and hexa and octa-
dioxin surrogates.  The samples were soxhlet extracted with toluene for  20 h and
the toluene extract concentrated to 3-5 ml and exchanged for  hexane.  This raw
extract was then subjected to clean-up on a series of three columns.  An  acid/base
silica column first removed the easily oxidizable organics. This was followed by a
silver nitrate silica column to eliminate sulphurous compounds.  The third column,
packed with activated  basic alumina was eluted with  two solvent mixtures  of
differing polarity to separate PCB and other interfering compounds from the dioxin
fraction.

Prior to GC/MS analysis the purified extract was concentrated just to dryness and a
performance standard solution  (20 pL  of 1,2,3,4-T4CDD  for  EC and 10 jiL  of
                                   623
-------
1,2,3,4,7,8-HeCDF for OME) was added. A summary of the instrument op
parameters for EC and OME is given in Table I!.
       TABLE II - SUMMARY of INSTRUMENT OPERATING PARAMETERS

Instrument
Column
Injection
Run Time
Mode
Oven Program
EC
Finnigan 4500 HR GC/LRM5
DB-530mx0.25mm ID
2 [iL, on-column or splitVsplitless
30min.
Electron Impact (El),
Multiple Ion Detection (MID)
100°C for 1 mm , to 180oC@30°, to
28QoC @ 4t>, Hold for 2 min.
Oft
Finnigan 4500 HR(
SE-52XL30mx0.2
2 |iL, on-column
25 min.
Electron Impact (E
Multiple Ion Detec
110°C for 2 mm., t
300°C @ 5". Hold i
Two ions of the molecular (M + ) cluster and an ion of the daughter  .
cluster were monitored for each dioxin/furan  homologue from tetra
Quantitation of dioxin/furan was achieved by comparing the response o
analytes in the sample to an external standard.

The presence of dioxin/furan in the purified sample extract was confirmed ,^)^.
of the following criteria were satisfied: (a) the value of total ion current*Bl-
each monitored ion exceeded background noise level by a minimum ratio   $.
(b) the peak area ratio of two ions within the molecular ion cluster T ondiJJj
homologous group were within ±25% of the ratio obtained for the corre!acter'JJ(i
standard component;  (c) the peak maxima for the three monitored char   ^
ions coincided within ± 1 scan unit; and (d) ions were detected within tn
retention time window for the homologue in question.                     J(

Additional quality assurance steps included the analysis of glassware
analysis of quantitation standards before and  after each batch of 5
establishment of a calibration  curve using  5 levels  of native
congeners) and analysis of NBS Reference Material (#1614).
A summary of the mean detection levels (surrogate recovery corrected) a.c^^e •
the three sampler types over the three active sampling days is given in  - -"
Penta dioxins could not be quantified in the EC samples because of an in1
problem.  Mean surrogate recoveries ranged from 70 to 79 % for the t
and from 69 to 92% for the OME samples. The standard deviations OT
mean recoveries were larger than the standard deviations of the
recoveries. All reported concentration data for each discrete sample were
for surrogate recovery.
                                   624
-------
     TABLE
             - MEAN DETECTION LEVELS  (pg/m3 - Recovery Corrected)
                  for THREE ACTIVE SAMPLING DAYS
OMOLOGUE
GROUP
VHM^HBBi^M
T4CDD
P5CDD
H6CDD
H7CDD
OCDD
••Mi^-«i^_iw^
T4CDF
P5CDF
H6CDF
H7CDF
OCDF
SAMPLER TYPE
EC HI-VOL
MEAN
(n = 12)
0.06
INT
0.10
0.10
0.25
0.05
0.06
0.08
0.09
0.20
OME HI-VOL
MEAN
(n = 12)
0.17
0.16
0.15
0.20
0.40
0.09
0.09
0 14
0.16
0.30
PS1
MEAN
(n = 12)
0.40
0.40
0.25
0.30
0.6S
0.20
0.20
020
0.20
0.45
L                      Results and Discussion
l&'ts forth
'ihl^ran  1  st Passive exposure samples (field blanks) are shown in Table IV.
      9e flowrates for each sampler were used to calculated an equivalent
     j °PCentration  value.  The passive filter sample from the OME  hi-vo!
     filt     'evel °^ contamination. The passive  hi-vol PUF sample and the PS1
       Ter and PUF samples for the  same day showed  no detectable
           >f any dioxin/furan homologue group. The passive fijter  and PUF
 ""lTorn  ""^ ^ ™-vol showed no detectable concentration of dioxins or furans,
   6 Pa^enta"^uran measured just above the detection level on the PUF  fraction.
 Centratio e samP|es collected after the  active sampling  days, no detectable
lo{?°Urce MnSL°^ dioxins or furans were found in any of the filter or PUF  samples.
  e$n id-    apparent contamination in the one OME passive filter sample has
^..   lc|entified.
C?
          active sampling days are shown in Table V. No data was available for
          sampler on Dec. 17 because  of equipment failure. For the three
  ^ol th1*5 n° detectable concentrations of tetra or penta dioxin were measured
     tne samplers.
V^p|jn   .
       u n^w ^1 results from the two EC  samplers agreed very well.  The other
   sarnnu   ° not match as well nor did each sampler type agree well with the
   °f dete types' although most positive results were within a factor of two or
   *f samilon levels-  Tne biggest discrepancy was between OME hi-vol A and
         5lers for tetra-furan  and between OME hi-vols A and  B and  other
        sforhexa-furan.
                                625
-------
             TABLE IV - DIOXiN/FURAN CONCENTRATIONS (pg/m3)
rajjivc joni\ji^3- uc^cini/ci i -», i Jt»/ __
HOMOLOGUE GROUP
T4CDD
P5CDD
H6CDD
H7CDD
OCDD
T4CDF
P5CDF
H6CDF
H7CDF
OCDF
EC HI-VOL
ND
ND
ND
ND
ND
ND
0.12(1, P)
ND
ND
ND
OME MODIFIED
HI VOL
ND
ND
ND
24.0(2,F)*
105(1,F>*
ND
ND
I.O(I.F)*
19.0(3,F)*
90(1, F)*
PS1
ND 	
ND^^^
ND^^^
ND_ 	 „
ND ^r-
ND_— -
ND 	
ND^— --
ND 	
ND 	 1
                             Suspect contamination
                                                                       I,


                                                                 a    °
There was even greater variability between samplers on sampling day #2
data interpretation is confused by the missing results for EC hi-vot A and
interferences in some of the samples which affected octa-dioxin and ne* j
results.  For this data set the EC sampler and the OME hi-vol A agree m9r^iy
than any other sampler combination.  The OME hi-vol B showed relati^ '1
concentrations of hexa and octa-dioxin and hexa and  hepta-furan 'n. l ^ fh*
fraction. All other samplers showed positive results only on the filter f ractio * j-l0t\v
positive PUF results are suspect since even the least volatile species (neP.u efilte'
and furan) were found on the PUF with no detectable concentration on tn  ^y-
The detection level of the PS1 samplers were relatively high on  this samPlJJ ^w
Four isomers of tetra-furan were detected by the EC sampler at concentrati
close to the detection level.                                               .
On day #3 there was again good correspondence between the EC s
though measured concentrations of all homologue groups except hepta a
dioxin were close to detection levels.  Results from the OME sampler pairs'
no detectable dioxins or furans except for a high concentration of tetra
PS1 sampler B. The lack of positive results for hepta and octa-dioxin for \
samplers is puzzling considering the estimated detection levels. Isomers^'  ^
other homologue groups quantified on the EC sampler were  at or °e
detection levels of the other samplers.
                                Conclusions

Dioxin and furan concentrations in Toronto ambient air appear to be clu'tleaStJ[f,|
               -                                            The     s "
hepta and octa-dioxins found in the  highest concentrations. The
concentrations are similar to those recently recorded in Windsor, O
substantially lower than measurements reported from Germany.3  For a
except one, dioxins and furans were found only on the filter fractions.  .
primarily due to the fact that the more volatile tetra and penta dioxins an
                                                                      '
                                   626
-------
       TABLE V- DIOXIN/FURAN CONCENTRATIONS (pg/m3)
                  on ACTIVE SAMPLING DAYS
    - December 16,1987
GROUP
^M
•••M

•^^M
^^•B
EC MODIFIED HI-VOL
A
0.12(1, F)*
0.48(1, F)
0.09(1,F)
ND
<0.16
800
B
<0.13
0.62 (1,F)
0.10(1, F)
ND
<0.16
872
OME MODIFIED HI-VOL
A
<0.10
0.60(1,F)
0.93 (1,P)
ND
0.7 (1,F)
662
B
<0.20
<0.40
<0 10
ND
0.4(1,F)
826
PS1
A
<0.10
<0.20
<0.03
ND
<0.06
363
B
<0.09
0.90{1,F)
<0.09
ND
<0.06
318
                                *T4CDD, PSCDD, H6CDD. P5CDF, H7CDF, OCDF not detected
•aJL£2_. Dec
GROUP
^/CDD
OCDD
"^CDF
^t>CDF
^CDF
•«»„_

ember 17. 1987
EC MODIFIED HI-VOL
A

--
--

—
--
--
Did not
operate

B
<0.10
0.83 (2,F)
1.88(1, F)
0.22 (4,F)
ND
<008
<0.09
831

OME MODIFIED HI-VOL
A
<0.08
0.30 (1,F)
1.40(1, F)
<0.02
ND
INT
<0.02
625

B
0.93 (1,P)
1.SO(2,P)
INT
<0.04
ND
1.3 (2,P)
0.93 (1,P)
826
PS1
A
<0.10
<0.60
INT
<030
ND
<0.20
<0.80
359
B
<0.30
<0.60
<0.30
<006
ND
<0.10
0.30(1,F)
335
*T4CDD,P5CDD, OCDF not detected
*»J£ijL December 51 1 QR7
ROup
5U?D "*
?CDD
^DD
^S 	
SCD^
6CDF
?CDF
JCdF ~~
^Mrr^)

EC MODIFIED HI-VOL
A
•^•^^^^^••B
0.78 (2,F)
1.46(1, F)
0.15(1, FP)
0.11 (1,F)
<0.09
0.04(1, F)
0.17(1,F)
800

B
0.08(1, F)
0.64 (2,F)
0,99(1, F)
0.04 (1.F)
0.04 (1,F)
0.05 (IF)
0.11 d.F)
<0.13
859

OME MODIFIED HI-VOL
A
<0.20
<0.10
<0.20
<0.03
<0.20
<0.20
<0.20
<0.20
625
B
<0.40
<0.10
<0.30
<0.10
<0.10
<0.30
<0.10
<030
722
PS1
A
<0.10
<0.50
<0.50
INT
<0.20
<0.50
<0.30
<0.30
392
B
<0.20
<0.20
<0.30
1 6(1, F)
<0.20
<0.30
<0.20
<0.30
343
* T4CDD, P5CDD not detected by any sampler
   erV detected on filter (F) or PUP (P))
                              627
-------
were measured below or near detection levels and  because of the
sampling temperatures (-5 to 2°C).  The sample for which positive results
found for the PUF fraction is suspect.
The Environment Canada sampler pair gave consistent results for the two
which  a complete  data set was available.  Penta-dioxins could  not be n
because of an interferent. Detection levels were consistently low (in the 0.0?*
pg/m3 range for tetra through hepta-dioxins/furans) and surrogate recover'6
high with  relative standard deviations of less than 25%.

Results from the OME samplers were more variable with no consistent <
between sampler pairs or between sampler types. The relatively high
levels for the PS1 samplers resulted in very few values over detection for any
dioxin/furan homologue groups. Based on the results of the passive filtersan
Dec. 14, the PUF sample for hi-vol B on Dec.  17 and the high tetra-furan  res
PS1 sampler B on Dec. 21, there may be a source of low level contamination w
influencing the OME results.                                              ,

It is apparent that additional work will have to be carried out to identify ^^
procedural differences that are resulting in the apparent inter-agency v~"
measured  ambient dioxin and furan concentrations.  When measured
furan concentrations are close to method detection levels, relatively
in results is unavoidable. Improvements in method precision  could undouuj  ^
obtained by resorting to high resolution mass spectrometry (HRMS). B.aseueV*'1
environmental significance of dioxin/furan at the detection  levels achievap  .,,,fi
LRMS, analysis by HRMS may not be justifiable. For example, with a sampI*
of 900 m3 and recoveries  of 80%, an ambient air concentration of 0.06
2,3,7,8-TCDD can be measured.

                                References

1.  Tashiro, C, Clement, R.E., Szakolcai,  A. and  Chan, W. (1987).
   Monitoring Techniques for Dioxins in Ambient Air". Proc. of 1987
   Transfer Conference, Toronto. Dec. 1, 1987.
2.  United States EPA (1986). "Compendium of Methods for the
   Toxic Organic Compounds in Ambient Air.  Method T09". Report
3.  Buck, M. and Kirschner, P. (1987).  "Imission Measurements of
   Dibenzo-p-Dioxinsand Dibenzofuran in North Rhein-Westfalia". State
   Westfalia Report TR87-0020.

4.  Liu, B.Y. et al (1983).  "Characteristics of air sampling filter media"-
   the Mining and Industrial Workplace, Volume III.  Ann Arbor Press.
5.  Environment Canada (1988). "Detroit Incinerator Monitoring
   Report #1:  Windsor Air Sampling Site".  River Road Environmental
   Centre, Ottawa, Ontario.
                                  628
-------
    r
 Of &Io?°UND ENVIRONMENTAL CONCENTRATIONS
    U*INS AND FR
 J,.

 Ho0| ***** and Ronald A. Kites
 r &etj9   blic and Environmental Affairs
 S»» IT ?Cnt of Chemistry
         , Indiana 47405
 fn   ^ete     °f analyzin8 ambient air for dioxins and furans in the fcmtogram per
 I,'*" con' rangc, nas been developed.  The  method is used to compare dioxin and
 (jN in Cc,ntrations in  air  collected at  three  locations  with  the concentrations
 'k   cd b*10 ant*  sed'ment'  These  comparisons  suggest  that atmospheric transport
  "Ugho^ Particle deposition  is a  primary  method of dispersing  these compounds
        the environment.
    >C in..
    iOSDh        n  °*  munic'Pal waste leads to the production and  emission  to
     OB  rc.of Polychlorinated dibenzo-p-dioxins (PCDD) and dibenzofurans (PCDF)
    dCp c.e ^ the atmosphere, these compounds can  travel long distances  before
  ^ent 0fltC(* 'nto t*ic env*roomcntJ this is demonstrated  by their presence  in the
  Cfic S   a remote lake whose only input of these compounds could be  from atmo-
       rCCS ^^'   To understand  these  transport  and depositional  processes,  it
        ' t0 Cxaminc tnese compounds  in each of  the various  compartments  so
        !°ns which  occur between  the sources and sinks can be  studied.   Previous
    vin«°USCd On c'tner combustion sources (1-10) or sinks such as sediment (77-
       « the transport medium, the atmosphere, to be  studied.
,   v

i^0ut)d$aye undertaken  a  study aimed at  determining the  concentration  of  these
V -'l  U th thc ambient air of Bloomington, Indiana.   Bloomington was chosen be-
li,lcil»3l Jhe  Proposed location  for  an incinerator designed to burn a  mixture of
   'tUdy a^te and polychlorinated  biphenyl  (PCB) contaminated materials.  Thus,
    Can KW'1'  Prov»dc baseline  atmospheric  concentrations of  these  compounds
    c win  USed  latcr to evaluate the performance  of the incinerator.  In addi-
         be able to improve our understanding of the atmospheric transport and
        Processes.
                                    629
-------
 Experimental

      Air has been sampled  at three  sites in Bloomington.  These sites 'i
 sites which  had the  highest  predicted particulate  concentrations from the
 tor  proposer's  computer model and one  site in the downtown area  about
 local population is centered.   All  sampling sites are located  on  roofs  of bu' rjsfli>
 which are 2-4  stories tall.  Sets of samples have also been taken  at two co^P  ^j
 locations.   These are a much  more rural site, Trout Lake, WI, and a more
 site, Indianapolis, IN.
                                                                          •  dra*fl
      Air has been sampled with a  modified high-volume air  sampler.  Air is   ^
 through a glass fiber filter and then a polyurethane foam (PUF) plug.  These ^ ^
 ials are analyzed  separately  giving a  determination  of particulate bound (*'  ^
 filter) and vapor phase (in the PUF  plug)  compounds.  It  is important to n°l  ej»
 vapor phase compounds could absorb to the filter or particulate bound compoun .^\
 be "blown-off"  the  particulate matter  or  pass  through  on   very  small  P8f
 thus, the two phases are operationally defined.
                                                                            \ef is
      Rain is sampled with a  one  square meter wet  only  collector.   The  s»|  P   -
 designed  so  that as  rainwater  is collected  it is pumped  through a  glass
 ter to create  operationally defined particulate bound and dissolved phases.

                                                                          114) ^
     The sample clean-up procedure is described more completely elsewhere i'   ac'
 consists of Soxhlet extraction followed by column  chromatography.  Prior to * g(1il
 tion, samples are spiked  with two carbon- 13 labeled isotopic  compounds (OC^  ^y
 1,2,3,7,8  PCDF) which are used  as  internal standards.   A  series of  external
 dards are used  to  measure response factors for each level  of chlorination
pared  to the internal  standard.  These external standards are also used in c  -^
tion with published retention indices  for  PCDD (75) and  PCDF  (16) to Sct  „, of
specific information for some of the congeners.  Only congeners containing '  ^
more chlorines  are  examined.   After clean-up,  the samples are  analyzed " ^
resolution gas chromatographic low resolution mass  spectrometry; the mass sp*   jofl
eter is operated  in  the  electron capture, negative  ion  mode  with  se'ecti   ^r
monitoring  (SIM).  It  is  important to note that  this  ionization  mode  g'vcs
response for 2,3,7,8-TCDD,

Results

     Previous work has shown that the chloro-homolog profile of dioxins *°  fr"1?
changes as  one  progresses from source to sink (12-13).  The profile chan?Cat«.
one in which no chloro-homolog dominates  to one in which OCDD is predofli»°,^  -.rf
it  is  assumed  that  the quantity  of combustion  sources  at a  close proxitf.
greatest for a large city (Indianapolis),  second for  a smaller city (Bit
and  smallest for a rural community (Trout Lake), then a  hypothetical tf8,n[flg '"'
chloro-homolog profiles with  distance from source  can be created by exam'0 cj,|0f?'
chloro-homolog profiles  at the  three  locations.  Figure 1  shows these  three  ^{v
homolog profiles.   Several important points are  noted: First, the  method  '  ^
ciently sensitive to determine concentrations in the  low femtograms per cu osph*f)j
range.   Second,  the vapor phase contains  a significant portion of  the ,8tB1D
-------
                    890 fg/m
   BLOOMINGTON AVERAGE AIR
   INDIANAPOLIS AVERAGE AIR

                 980 fg/m3
                      86 pg/L
         AVERAGE RAIN
                 890  fg/m


BLOOMINGTON AVERAGE AIR
                     970 pg/g
AVERAGE GREAT LAKES SEDIMENT
                  160 fg/m

TROUT LAKE AVERAGE AIR
Figure I.  (left)    Chloro-homolog profiles for average air  from Indianapolis, IN;
Bloomington, IN; and Trout Lake, Wl.  The total OCDD concentration is given as a
scaling factor.

Figure 2.   (right)     Chloro-homolog  profiles for  average Bloomington  air;
Bloomington rain; and Great Lakes Sediment. The total  OCDD concentration is given
as a scaling factor.
                                    631
-------
                                                                        j
                                                                     n
material suggests  that particle deposition (wet or dry) is  the  primary dep°
mode of these compounds from the atmosphere.

     To further  examine depositional processes the  flux to the sediment
previously (12) is compared to the flux implied by the measured air conce
The  implied flux  is determined by multiplying the air concentration by  the   ,
tional velocity and  proper conversion factors.  Other  work has given  an **  ^
depositional velocity of  1 cm/sec for organic compounds (17).  This data COB*
is  given in Table I for fluxes  implied  by either  the  total  or particulate
only concentration.  If the two rural sites arc compared, it is seen that tn
Lake flux  is less than  Siskiwit  Lake  but  it  should  be  noted that  Tft.u." it
Ontario (a  city of  100,000) is approximately  40  miles  Northwest  of S
The  average Great Lakes flux  lies somewhere  between  Indianapolis  and
which seems correct as  there are both urban  and rural areas in  the
area.
      Table l.   Comparison  of measured  and  implied  flux**'


              PCDD &  PCDF  CONCENTRATION         IMPLIED FLUX

                     TOTAL   PART.  ONLY        TOTAL     PART. Ol*

                                                 pg/cm2yr    pg/clD ^

                                                     17           9.0

                                                     70          51

                                                   165         130

TROUT LAKE
BLOOMINGTON
INDIANAPOLIS
pg/m3
0.53
2.2
5.2
pg/m3
0.28
1.6
4.1
MEASURED AVERAGE GREAT  LAKES SEDIMENT  FLUX  =120

MEASURED SISKIWIT  LAKE  FLUX -  20 pg/cm2yr
	-^'

                                                                     5 « <'
     The data shown suggests the following scenario.  Combustion  Pro<*uC j,c A
variety of chloro-homologs.  As these compounds are transported through   $ '^
sphere, a degradation  process occurs on  the vapor  phase.   This process re   jo<
total air  concentration,  but  significant amounts remain in  the  vapor  ph*  ngft'Lj
equilibrium with particulate bound  material.  This equilibrium favors *  .tfrgO J
late bound material for  the higher chlorinated homologs, and thus they u° .e(t ^
degradation.  Eventually,  the particulate matter  deposits  from  the atmf
-------
  J)
 tig' *• Bumb, W. B. Crumett,  S. S. Cutic, J.  R.  Gledhill, R.  H. Hummel,  R.  P.
 l|{'L- L. Lamparski, E. V.  Luoma, D.  L. Miller, J.  J. Nestrick, L. A. Shadoff,
      l, and J. S. Woods, Science 210, 385-390 (1980).
   U
 (ISW1 R. Buser and  H.  P.  Bosshardt, Mitt.  Cibiete  Lebensm. Hyg.  69,  191-199

 !,
     ^avallaro» L. Luciani, G.  Ceruni, I.  Rucchi, G.  Ivernizzi,  and  A.  Garin,
         11,859-868 (1982).
  l( ...
 ^S* .  berti  and D. Brocco, Chlorinated Dioxins  and Related Compounds:  Impact on
 ^•ISi  nrnen^  (Eds. O.   Hutzinger, R. W. Frci, E. Mcrian, and F.  Pocchiari),  pp.
    '  pergamon Press (1982).
 *  4
 ^>' ^-'berti, D. Brocco, A. Ccciata,  and  A.  Natalucci, Analytical  Techniques  in
    Cental Chemistry, (Ed. J. Albaiges), pp. 281-286. Pergamon Press (1982).
   j
 ^OJ ^' A.  Lustenhouwer, K.  Olie,  and O.  Hutzinger, Chemosphere 9,  501-522

 \
    °lie' J< W* A- Lustenhouwer,  and O. Hutzinger, Chlorinated  Dioxins  and Re-
            .'   Impact on  the  Environment. (Eds. O.  Hutzinger,  R. W.  Frei, E.
.         F. Pocchiari), pp. 227-244. Pergamon Press (1982).
  C b
!S*o?Ppc> S< Marklund, P. A. Bergqvist, and M  Hansson, Chlorinated Dioxins and
\el    ns  in the  Total  Environment,  (Eds.  G. Choudhry, L.  H. Keith,  and C.
.    ;'DP. 99-124.  Butterworth (1983).
•H.VT
     • rong and F. W. Karasek, Chemosphere 15,  1219-1224  (1986).

    ' Rappe> s> Marklund,  L.  Kjeller,  P.  Bergqvist, and  M  Hansson, Chlorinated
     and Dibenzofurans  in  the  Total  Environment  II,  (Eds.  G.  Choudhry, L. H.
(|   ' nd C Rappe), pp. 401-424. Butterworth (1985).

  
-------
CONGENER PROFILES OF POLYCHLORINATED DIBENZO-P-DIOXINS
AND DIBENZOFURANS IN ATMOSPHERIC SAMPLES
Jean M. Czuczwa and Sylvia A. Edgertona
Battelle Columbus Division
505 King Avenue
Columbus, Ohio 43221
                                                                   nee 
-------
Auction

      Tl?e   atmosphere  appears   to  be   an   important   reservoir  of
      rinated   dibenzo-p-dioxins   (PCOD)   and  dibenzofurans   (PCDF).
     er^c deposition of PCDD/PCDF on land or water  surfaces  may be one
         p.CDD/pCD.F  detected at trace  levels  in environmental  matrices
        air, sediments, water,  and biological  tissues.

      Possible   sources   of  atmospheric   PCDD  and   PCDF  include  the
Pfi1ta Lturei   use«   an£l   disposal   of   technical    products   such   as
%ni  or°Pneno!  aP" P^B  anc'  combustion sources  such as  municipal  and
   ij *^  waste incinerators,  kraft  paper  mill  boilers,  and  automobile
         Because of the lack  of expected chemical  production  sources in
      mbustion sources were thought to  be  the  dominant  local  sources of
      F,
           aPProach  to  determining  possible  sources  of  atmospheric
           \° comPa*"e  congener profiles  of the atmospheric samples with
for   °   possible sources.  We  present here  congener  profiles determined
to ^dn)ples  from  four locations in Ohio.  Two sampling  sites  were chosen
fyl ! close  to and downwind  of municipal  solid waste  (MSW)  incinerators
to ,. ]J*a9e sludge incinerators  (SSI).  A third  site was  located adjacent
]eyei  van highway traffic  (a  site  chosen  for  its   high  measured  lead
'icer •  w^e  a  f°urth site  (remote)  was located away  from  any obvious
tossikn on  sources.    Comparisons  of the  air  congener  profiles  with
   1D'e  sources  of PCDD and  PCDF are made.
p

 Per1"iental Methods

v°liim  Slx  air samples  were collected  in  November,  1987 using  a  medium
Poly'* air  sampler (~ 0.3  m3/min) fitted with  a  glass fiber  filter and
f°Hr   rttlane ^oam P^U9*   Total  volumes  of 800-3000  n»3 were  collected.
  r-jp  t^11e  polyurethane  foam  plugs were spiked with 10 ng  of  1,2,3,4-
     ?   3ci2 Just  Pn?r to  sample collection.   Sample  locations  and
          source  locations   are  shown  in  Figure 1.    Possible  local
     erc sources of PCDD/PCDF at each location are  given  in Table 1.

CSoi .^xtraction*  column  chromatographic  separation  of  cpextracted
%0  "Jds and  determination of  PCDD and PCDF by  combined capillary  gas
(3)%ndtography/high resolution  mass spectrometry are described  elsewhere
J^e  n    field  blanks, a laboratory blank,  and  a  spiked blank  sample
'dentp^ePared with the air samples.   The QA/QC criteria  that were used to
  11 fV  PCDD and PCDF isomers  included:

    '  Simultaneous responses at  both  ion masses;
  (p\
    '  Chlorine  isotope ratio within +/- 15% of the theoretical  value;

    '  Chromatographic  retention  times  within  windows  determined  from
      analyses of standard mixtures;

    '  Signal-to-noise  ratio  equal to or greater  than 2.5 to  1
      for both quantification  ions.
                                 635
-------
                                                                  /  l'
        The  recoveries  of the  internal  standards averaged  100 +/• #,
These recoveries indicate excellent control of the analytical proceflju ^
The  recoveries  of  the  1,2,3,4-tetra-CDD-13Ci2  spiked on  four  ot^
polyurethane   foam   plugs  before  collection   averaged   84  +/-  T1)
suggesting  only minor  losses  of the  lower chlorinated  (more
PCDD/PCDF.
Results and Discussion

        Total  PCDD/PCDF  measured  in  the  Ohio  samples   are  giv6"
Table 1.   Detailed,  isomer-specific PCDD/PCDF  data  for these sites
be  presented elsewhere  (4).    The  total  PCDD/PCDF  concentration
from  1900-9900  fg/n)3.    The  lowest detected  PCDD/PCDF  levels were
for the sample collected at an  urban highway  traffic site  (located
intersection  of  17th   Street   and  Highway   1-71  in  Columbus).
PCDD/PCDF  are higher  than  those  reported for  Bloomington,  Indiana
Trout Lake,  Wisconsin  (1) but  were 5-10 times lower than  levels meas
in densely industrialized locations in West Germany  (2).
                                                                     vt
        No 2,3,7,8-tetra-CDD were  detected in the Ohio air samples. ?
average  detection limits  of  less than  240  fg/m3.    However,  2,
tetra-CDF  (130-490 fg/m3)  and high  concentrations (890-3800  f  '
total  tetra-CDF  were   found  in  all   samples except  the  urban
traffic  site.   It should  be  noted  that  2,3,7,8-tetra-CDF may
completely resolved  by HRGC from  other isomers  in  the tetra-CDF
and  thus,  may  contain  contributions  from  other  isomers.    Low
(< 100  fg/m3)  of  2,3,7,8-substituted  hexa-CDD,  penta-CDF,  and
were detected.

        A  PCDD/PCDF  congener   profile   (relative  concentrations  ot
tetra-  through  octachloro-CDD/CDF) for  a  typical air  sample,  Wal«o
shown  in  bar chart form in  Figure  2.   No  tetra-  or  penta-CDD  <$
detected in  the Waldo  sample.   The congener profiles for  the air ^J-aii
showed  certain  trends.   In  general,  the  concentration of dibenzot  ,0,
decreased  with  increasing level of chlorination while the concentr  p
of  dibenzo-p-dioxins   increased with  increasing  level  of chlorin3
Tetra-CDF was often the most abundant PCDD/PCDF  congener.
        These  congener profiles  show  similarities to  those __     t,.
Indiana  (1)  and West Germany  (2)  and may  be  related  to  combu>.les
sources.  A literature search was conducted to develop congener Pr
-------
   ?• 0. Eitzer, R. A. Hites, "Dioxins and  furans  in  the ambient air: A
            study,  Chemosphere.  in press.
2.  r  ft
   *• Rappe,  L-0.  Kjeller,  P.  Bruckman,  K-H.  Hackhe,  "Identification and
   Quantification  of  PCDDs  and  PCDFs in  urban  air,  Chemosphere  17:  3.
   U988) .

   5: A. Edgerton, J.  M.  Czuczwa, "Source Apportionment of  Dioxins and
   "Tbenzofurans  in   Ambient  Air  in  Ohio,"   presented  at  the  APCA
   International Specialty Conference, "Receptor Models in  Air Resources
   Mar»agement,"  February,  1988.
4,  c
   G* A-  Edgerton,  J. M.  Czuczwa,  J.  D.  Rench,  D.  A.  Egan,  R.  F.
   "oaanbosi,  P. J. Koval,  "Determination  of Polychlprinated Dibenzo-p-
   ^oxins  and Dibenzofurans and Associated  Health Risks in  Ambient Air
   2n Ohio,"  submitted for presentation  at the 81st Annual  APCA Meeting,
   PaPer No.  88-77.1,  May,  1988.

    * C. Siebert,  D.  R. Alston,  J.  F. Walsh, K. H. Jones,  "Statistical
       erties   of   Available  Worldwide  MSW  Combustion  Dioxin/Furan
       sions,"  presented at the 80th  Annual  APCA Meeting,  Paper 87-94.1
       , 1987.
6,  „  c
   ;;• S. Environmental  Protection Agency,  "National Dioxin  Study Tier 4-
   Jjjbustion  Sources,  Project  Summary  Report,"     EPA-450/4-84-014g,
   (lq1ce  of  Air   Quality   antj  Planning,  Research  Triangle  Park,   NC,
         was  supposed  in part  by the  Ohio Air  Quality  Development
           We  would also  like to  acknowledge Bob  Hodanbosi  and  Paul
     °f the Ohio  EPA  for  their  direction and assistance.
                                 637
-------
       Table 1. PCDD/PCDF in Ohio Air Samples
                         Location
                                     Total
                                 PCDD/PCDF
                                    (FG/M3)
       Municipal
       Incinerator
       Urban
           Akron (2 km)              5,600
           Akron (Colocated)         6,400
           Columbus (1 km) 11/18      4,600
           Columbus 11/20            9,900

           Highway Traffic            1,900
       Rural
           Waldo
3,600
                                      ^Cleveland
                                       •          o
                     &I-
              Springfield^+ColumbuS
       0 Middletown

   g Hamilton
Figure 1.
                                  + Sampling

                                  • MSW

                                  • SSI

                                  ° HWI

Map of Ohio showing sampling locations and potential P |
sources.  The Waldo site is near Marion, Ohio.  MSW s *
solid waste incinerator; SSI * sewage sludge incinerat  »
hazardous waste incinerator.

                   638
-------
                                                Municipal Incinerator and Waldo
05
CO
CO
                               0.25
                               0.20
                               0.15
                               0.10
                               0.06
                               0.00
                                                                             Municipal
                                                                             Incinerator
                                       2378  TCDF  PCDF  HXCDF HPCDF OCDF  2378  TCOD  PCDD  HXCDD HPCDD  OCDD
                                       TDCF                              TCDD
                               Figure 2.  PCDD/PCDF congener profiles of the Waldo air  sample measured  in this study
                                         and an average  congener profile  for municipal incinerator emissions.   The
                                         average municipal incinerator profile was calculated from literature values
                                         as described in reference 3.
-------
DIELECTROPHORESIS OF CHLOROPLASTS-
A  NEW TECHNIQUE IN BIOMONITORING OF
LOW LEVELS OF SO0
Adeel Ahmed*,  M.S.R.Murthy+,  S.H.Raza*
and G.Gopalakrishna*
*Biophysics  Laboratory,  Nizam  College,
Hyderabad-500 007, INDIA
^Department  of  Botany,  Osmania University,
Hyderabad-500 007, INDIA.
      Dielectrophoresis   is   the  migration  of  neutral  particles
a  nonuniform  electric  field  (a.c.   or  d.c.)   towards  the  re8 ^g
of  highest  field  intensity.   Dielectrophoresis  of  chloroplasts
measured  interms of collection of  chloroplasts on  electrodes occu jo-
in  the  form  of a  chain in  a unit  time and  it  is called  'D gts
trophoretic  Collection1.   Dielectrophoretic  mobility  of  chloropl3
as  a  function  of frequency  of  the applied  electric  field  (0.1M
1.5MHZ)  is  found  different  in  plants growing  in  different  enVlf gts
mental conditions.  Further dieletrophoretic collection of chloropl
was  less   in  polluted  area  plants  than  in  control  plants  and f
found  to   decrease  with  increased  dosage  of   S02.  The  techniQ
could be  used  to monitor low levels  of S0.
                                  640
-------
                           INTRODUCTION

 Ca  Dielectrophoresis  is  the  translational  motion  of  neutral  matter
 (a cec*  by  the  polarization  effects  in a  non  uniform  electric  field
 by' •  or  d.c.).   It  differs  from  electrophoresis  - the  motion caused
 *heth   resP°nse  to  a  free  change  on  a  body  in al  electric  field
 e'ect     un^orm   or  nonuniform.  Peril and   his  colloborators  made di-
 ^  °Phoretic  analysis  on  biologi'cal  matter  at  cellular,  particulate
 ejec m°lecular levels  (1).  At  different  frequency levels of the applied
 tersLric fields,   living cells or  organelles  exhibit  specific  and chara-
 antjric  dielectrophoretic  response  and  hence  and  the  biochemical
    Physiological  state  of the  biological  system  could   be  analysed.


 In  Prolonged  exposure  of low  levels  of  S02  brings  invisible  effects
 Of g ^nts  such as gradual  stunting, decreased  leaf area,  high levels
 etc (.^ ate  content,  decreased  chlorophyll,  altered  transpiration  rates
 Sij^12-3)  and  these have been often used for  the  monitoring  purposes.
 thy   chloroplasts  are  the  primary  targets  of   S02  toxicity,  the
 fyg *°l°gy and biochemistry of  chloroplast would  generally be effected
   ^ the low  levels of S02  and they do  exhibit  effects very imme-
        However  studies  on  S0~  toxicity on  chloroplasts  are  very
 -Cig      Hence  an  attempt  has  been  made   in the   investigation  to
 Of e the  dielectrophoretic  collection  of chloroplasts  with  the degree
 the potion  and  to  characterise  the   immediate damages caused  with
    requency  peaks  and  shifts of dielectrophoretic  collection.


                    MATERIALS  AND METHODS

   Dielectrophoresis  was  done under  spherical field geometry  using
 .u ,/~pin  type  electrode  figuration.   A  pair  of   platinum  wires  of
 't) s  diameter  was  placed 1  mm  above  the  surface of  a glass  slide
 theirch  a  way that  their axis  lie along  the same  straight  line  with
     rounded  tips  facing each  other and   were  seperated  by  350
    18•  The  wires  were  passed  through   a  nonconducting ring which
      capacity   of  0.3  ml and  1.0 cm internal diameter.  When  this
    ls  cemented on  a  glass  slide   it  forms  the pin-pin  electrode
     r- The electrode chamber was  mounted on  a conventional  micro-
      stage  and  the  observations  were  made  with  an  eye  piece
 -.si eter  worked  in  to  10  microns  at  10  x  of  the objective.  The
%r 8^als  from  Radar  signal  generator  were  fed  to the  electrodes
°1iCs °°taining the required amplification. Conductivity meter (systr-
"V ~   )•  spectro  colorimeter  (systronics-103)  were  used  to deter-
S%en°ncluctivitv  and concentration (OD at 670  nm) of chloroplast
    si°n respectively,
        samPies  of Amaranthus  vidii were  collected  from  control
          gardens),  industrial   areas  (Nacharam)   and  S02  treated
       in  earlV  Corning  hours  of  the  day and  were subjected  to
     n  of  chloroplasts  using  Ting,  et  al.  (4).  The  mean and  peak
                                641
-------
                                                     3               3
levels  of  S02   in  industrial  areas  were  30  ug/m and   125  ug
respectively.  Plants  were  treated  with  SCL  for  0.1  and  0.01  P ^
concentrations   for  one  week  with  10  hrs  daily  treatment  Pe  f
using  standard   methods   (2).  Leaf  samples  were  also  analysed
chlorophyll  (5)  sulfate accumulation  (4)  and  relative  water con
(7).

      The  chloroplasts  were  suspended in isotonic glycine  and g
(1.5%  glycine   and  4%  glucose in  the volume  ratio  9:l)and a
volume  (0.2  ml)  was  dropped  in  to the  chamber  and  the £
were  applied   between  platinum   electrodes.   The  chloroplasts   v°^
collected  at  the  rounded  tips  of  the  electrodes in the form  of  P  {
chains. The average  chain  length was  measured  for 2 minutes ullgtg
microscope  which gives the  dielectrophoretic  collection.  Chlorop   .
were  kept  on   ice  prior  to  dielectrophoretic experiment  and  exp
ment  was  carried out at  room temperature  (26°C).  The  dielectrop   ,
retic response  of chloroplasts as  a function  of frequency  at  a  l*  „
voltage, the  electric  field  exposure  time  concentration and  con
ivity of the suspension
                             RESULTS
                                                                functi"
      The  dielectrophoretic  collection  of  chloroplasts as  a  IU  ji
of frequency  in Amaranthus  virdis at  different  environmental c°
tions  is  depicted  in  Fig.l.  In  all  plants  the  behaviour  ip    S0
rather  non-linear.  In  control  plants  there  is  a  continuous  i^c
in collection  of chloroplasts from  0.1  to  1.2  MHZ beyond  that
fall  is  observed.  Similar  behaviour  was  noticed  in plants  co*  t
from  Nacharam  industrial  area  however  the  quantum  of  coUe  ,^
was  less.  In  plants  treated  with SCL gas,  chloroplast  show  a  3  j.
varied  behaviour.   The  dielectrophoretic  collection   is   very   ^.^
at high concentration.   At  low  concentrations,  the  frequency _ Pe.gji
found at  1  MHZ  followed  by steady  decrease.  In  contrast,  *n  *$$
concentration   treated  plants  two  frequency  peaks 0.8  and  1-2
are noticed.


                             DISCUSSION

      The   physiological  state  of   chloroplasts  under   SO^  Poll?city
is much  altered  as  these  are  the  primary   targets of  S02
for the damage.  In  general absorption  of  SCL results  in  the g
tion  of  toxic  free  radicals  viz.SCT.,  CL,  HSCL,  OH  and  S04
These radicals bring  changes in membrane  permeability   (9), chl°r
decline   (10),  sulfate  accumulation  and  water  loss  (11).  1°
investigation  also  it  was  found  that  there   was  a  20.25%
in chlorophyll and water content  and a  3  fold Increase sulfate
mulation than  in  control.  These  factors  would  definitely  alter
physiological  biochomical state  of  chloroplasts.  Such a phy
change  would  ultimately   alter  the  polarization  characters
to different  dielectrophoretic  behaviour.


      The   polarization  changes  due  to , physiological   decline
characterised  with the  frequency  peaks  and  shifts of
retic  collection of  chloroplasts.   In  plants  growing under

                                  642
-------
les ^ions  dielectrophoretic collection of  chloroplasts  was  relatively
Cau .with a  peak  at  1.2 MHZ.  It is  due to the  altered  polarizability
of  6c^ because  of  loss  of  polar substances like water  and accumulation
Inte ?4'  T^e  P^k at  this level  is associated  with  Maxwell-wagrier
C0l r'acial  bulk  polarization  probably modified  by  surface  associated
the ?ctivity  which  can  operate  on  the  material  enclosed  with  in
   ^dividual  thylakoids.
ff   *i  SC"   treatment  plants  there  was  a  backward  shift   in  the
Ttjj  ency  maxima  followed  by  decreased  collection  of  chloroplasts.
     trend   gains  support  from  studies  of  wiley  (12)   on  canine
      Cytes   wnere   it  was  reported   that  there  was  a  decline  in
    jPcytes  collection followed by shift in  the  frequency  from  1MHZ
shift   due  to  sodium  cyanide,  a  respiratory inhibitor.  The  frequency
of  s  and decreased  collection mainly  must  have  resulted from  loss
^fin       sulfate  accumulation  chlorophyll  reduction.  These  factors
the 8s  ionic  imbalances  in the chloroplasts  and  could effect  or  alter
sPe Conductivity of  stromal  matrix   there      by  changing the  yield
freQu     at  surface  conductance  modified Maxwell-wagner  polarization
Qtl  chn°y ranges-  It  has  been  also  reported  in  serveral  experiments
stg^oroplasts  (4),   on  canine erythrcocytes  (12)  that  the  relative
'0rUc    tion  of  dielectrophoretic  collection  was  due  to  halting  of
the •  losses   or to   the  maintenance   of  stable  ionic  environment  in
i    *?terior  of  the cells.  Further  at  higher dosages the yield  spectra
    further  very  poor.  It  reveals that a chracteristic  relationship
   een  dielectrophoretic  collection and S02  pollution can be  drawn.
                           REFERENCES


2,  [J'A.  Pohl and  J.S. Crane 1971.  Biophys. ,J.  11,  711.
   JV N.Rao  and  F.Le  Blanc,  1966.  Effects  of  sulfurdioxide  on  the
   iichen  alga,   with  special   reference  to  chlorophyll.  Bryologist
3,  £9- 69-75
   D-w- Cowling  and   D.R.Lockyer  1976.   Growth  of  perennial   rye
   8rass  (Lolium  perenee  L.)  exposed   to  a  low  concentrations  of
4.  !ulfurdioxide. J_.  Exp. Botany.  27,  98: 411:17.
   **p-Ting.  K.Jolley,  C.A.  Beasley,  H. A. Pohl.  1971.  Dielectropho-
5,  [!ests  of  chloroplasts.   Biochem.  Biophys.  Acta.  234:   324-329.
   u-I.Arnon  1949.  Copper  enzymes in  isolated  chloroplasts,   poly-
6.  Phenol  oxidase  in Beta vulgaris. Plant  Physiol.  24(1): 1-15,
     18 • Patterson  1958.   Sulfur  In  :   Colorimetric  determination  of
   J0rimetals.  International  Science  Public.  Inc.,  NY,  pp. 216-308.
   '/•Singh   1977.   Practical   Plant  Physiology.  Kalyani  publishers
         Delhi)p.266.
       Peiser,   F.Shan and   Yang  1977.   Chlorophyll  destruction  by
       bisulfite-oxygen system.  Plant  Physiol  60:277-281.
              1987  Ph.D.  Thesis, Osmania  University, India.
           ,   1970.  Effects  of  atmospheric  S09  on  plants.  Sulfur
U, -institute.  J.  6,  57.                           i
    •K.Singh  and   D.N.Rao,  1983.  Evaluation  of  plants  for  their
    °lerance  to  air pollution.  Proc .  Symp ^n  Air  pollution  control
ll  eld at IIT,  Delhi,  Nov. 1983  pp. 218-224.
    • Wiley ,   1970.   Dielectrophoretic   studies  on  erythrocytes  and
lav
       obacteria.  M.S.  Thesis.  OK  lahoma State University.

                                643
-------
                                             _  30




                                             c  25
                                                20
                                              o
                                              £
                                              §  10
                                              5  5
O5
                                                 30
                                                          Fig£. Dielectrophoretic coWection of Amoronthus  yirdis chloroptasts

                                                                                 a s a functi on of frequency
                                                      .2
                                                                  Control
                                                           .t,    .6
                                                                             12  1A   '-6
                                                              0-01ppm SO 2 treatment
3S




30




25




20




15




10




 5
 30:




 25




 20



 15




 to



  5
                                                                                                                   ponuted area
                                                                                                                                   L2   IA   1.6
                                                                                                                  C)jppriSO2 treatment
                                                                                                                                         \f.   \&
-------
        AND CONSIDERATIONS
   T0 AIRBORNE ASBESTOS SAMPLING
  OUTDOOR  ENVIRONMENT
          Consultants
    California
        g  ambient  airborne  asbestos  concentrations  in  the  outdoor
       Is  useful  for a  number of  reasons,  but is  different  in many
    n,  indoor  monitoring.    Factors which  must be  considered
   °f  an  outdoor  airborne  asbestos  sampling  include  the  choice  an
   *te sample volume, sample  duration,  the lower detection limit,
    l°«Mng  and total paniculate  matter loading on  the  fll
kii°f  samples  necessary  to achieve  the  desired  objective,  and the
  l1*ies of the sampling equipment.  Potential  problems which may  ar se
  *  the   difficulty   or  impossibility   of   measuring  low  asbestos
  isions  in  areas with  high  ambient TSP, the  lack  of  the necessary
  ^uired  to  choose  the  appropriate  sample  volumes  and  number
   >  and  physical  limitations  of  the  sampling  equipment.
    win  allow  the Investigator to circumvent most of these potential
   i
                                645
-------
INTRODUCTION
                                                                       It
    In order to manage the  risk  posed by asbestos in our environment!
necessary to identify areas  and  activities  that lead to high exposures
then  reduce  the  exposures,  generally by  removing  or  containing
asbestos.   Techniques  have  been  developed  for measuring  airborne  asi
concentrations.   These  methods  work  well   in  most  cases  in  the   ent.
environment, but  are  often not well  suited to  the outdoor  enviro    QS
However,  there  are   several   reasons  for  measuring   airborne   a=
concentrations  outdoors.   Such data can provide  a background a9a'n!L
to compare  measured  indoor concentrations.    They  also can provide  tn
needed to assess the health risk posed by asbestos  in  the  outdoor air.

    Outdoor airborne  asbestos  measurement  programs have  been  sP°.nS^inent^
carried  out  primarily  by  federal  and  state  health  and  env'^flatio'15
agencies.   Most of  these  studies  have  been site-specific  investig   ^
associated  with  a known  or suspected  source of  asbestos, altn°.u.^Hon  °
have undertaken  to  characterize  the  spatial  variation or  district
airborne asbestos concentrations.
                                                                     Or an
    This  paper will  present  some  of  the   differences  between in,°ratio(lS
outdoor airborne asbestos  sampling, and examine  some  of the consia6
and problems related to outdoor sampling.

Differences Between Indoor and Outdoor Airborne Asbestos Monitojlll
    Airborne  asbestos  sampling  in  the  outdoor  environment
indoor sampling  in a number  of  ways.   In general, the  purpose
sampling  is different  from  indoor  sampling.  Indoor  sampling is
conducted  for  the  purpose  of answering one  of three  specific  <•
1)   Are  occupational exposure standards being exceeded?  2)   Are
asbestos  levels  high enough to require  an abatement action? or 3)   .
abatement  action  successfully  reduced  the  elevated  concentratio^ ^
purposes  of outdoor  sampling, on  the  other hand, are  varied.
desired  to characterize spatial or  temporal patterns,  estimate
locate sources,  or  assess transport  patterns.  Consequently, the
may take  the  form of time-series  sampling,  concurrent sampling at
locations,  randomized  sampling  in  space and/or  time,  or  upwind/
sampling.                                                                .
                                                                   • It/ W  r,fl
    The concentrations  of  interest in  indoor  sampling are reasons0  ^ w
defined.   Release criteria for  abatement contractors are  typical'J^gj  *
order  of  0.01  asbestos  structures/cm-3.    There  is  usually n°they  *(
accurately   quantify   significantly   higher   levels,   because  *  oU^s
indicative  of a site  which  will  require  further  cleanup.     J  ast>eS*i1
airborne  asbestos  studies,  concentrations-from  less  than 0.00I      fl
structures/cm-3   to  several  structures/cm13  have  been  measure0-   it
concentrations are  of interest,  and  it  is usually desired to  rtlian
as precisely as possible.
                                    646
-------
conjje  sources  of asbestos  which contribute  to  outdoor  asbestos  differ
      d^y  from  ^ndoor  sources.   While  indoor  asbestos  is  generally
         ^° the Presence °f friable asbestos-containing material, outdoor
    °s may  be  emitted from  vehicle brakes and  clutches, waste disposal
      Wind  erosi°n  °f  natural   serpentinite   soils,   resuspension  of
      -containing fugitive  dust, or the few manufacturing processes which
        asbestos.
In J"16 equipment used  for  sampling of outdoor  airborne  asbestos differs,
       cases»  from  indoor  sampling  equipment.    While  indoor sampling
        involves  the  use of a  110-volt pump, outdoor  sampling  is often
        with 6- or 12-volt pumps using  internal or external batteries and
        timers,  or a  110-volt  pump  with  a  generator.    Often,  outdoor
       must  be left  unattended.    If many  samples  are to  be collected
               the  sampling  apparatus  must  be  able to  start and  stop
  On  .—..j.    If  secure  locations cannot be found,  the  samplers  must be
   sP1cuous  in  order to avoid theft  or vandalism.

'V41laK?Jor difference between indoor  and outdoor  asbestos sampling is the
% h  Je guidance  for  the  design  and  protocol  of sampling  experiments.
    •*•  Environmental  Protection  Agency has published  several documents
     Provide detailed  procedures  for indoor air sampling.  The author is
    of no  such  guidance for outdoor  airborne  asbestos sampling.

   DERATIONS  IN THE DESIGN  OF AN OUTDOOR MONITORING PROGRAM

          Analytical  Procedures
           asbestos is commonly  measured by filtering a  known volume of
      a1r through  either a cellulose  ester or polycarbonate  filter.   A
    h  °f the f^ter is then appropriately prepared and examined in either
    nsm1ssion  electron microscope  (TEM)  or a  phase-contrast  microscope
   hL,SamPle preparation for  TEM analysis is performed by carbon-coating
   SU  Ulate matter  on  the filter and  then dissolving the  filter away.
t^osr     f1lm and  paniculate matter  are then  transferred to an election
     °Pe grid.   The TEM  method  is much more expensive than  PCM,  but has
            of  comparatively excellent resolution.  Fibers with diameters
           0.01 ym  may be detected by TEM.  Sampling and analysis methods
        have  been  Presented  by  Yamate  et  al.  (1984), NIOSH  (1986)  and
      and  Rood  (1983).    Analysis  by  TEM allows  confirmation  of  the
  >apM1ne  structure of the observed fibers using Selective Area Electron
    cI on   (SAED)   further  confirmation  of  asbestos  using   Energy  -
    S1ve Spectroscopy  (EDS).

     the TEM analysis, asbestos  "structures"  (bundles,  fibers, clusters,
    n    c°unted and measured  in  each  of a  number of  defined areas known
      °Penings.  Each  grid  opening  is  roughly 0.08 ym  square.
                                  647
-------
                                                                       116'"
    The PCM method Is much  less  expensive,  but  cannot detect fibers sinj^
than about  0.25  vm in  diameter.   Because  outdoor distributions or °  ^
airborne  asbestos  in the  U.S.  generally  have relatively  few flDfTte for
diameters  as  large  as  0.25 ym ,  PCM  is generally  not  aPPr°Pr^eS ««*
analysis  of  outdoor  airborne  asbestos   samples.    The  method  Q°  Q^
positively  identify  the  observed  material  as  asbestos,  and  ma™  afl b«
fibrous particles may  be counted.   For  this reason,  PCM  analyses
considered as only an index of asbestos concentrations.
                                                                       h fl0
    A third analytical method which  has  found limited use,  and for «
widely-circulated protocol  exists,   is Scanning Electron M1crosc°^w
The SEM method allows  elemental  analysis  by  EDS, but does not  ai '"1
The resolution  is  only slightly better  than  PCM, yet SEM  is  sig"1T
more expensive (EPA 1985).                                                ^

Sample Duration.   The  sample volume is  the product of sample  dura^a
flow rate.    In  many cases,  there  is  a  need to  sample  over  a gi» .  ^H
period  (such  as  during  the  period  of  one  direction  of a  diurn ^
pattern,  or  during working  hours).   In these cases, there  is  litllduratiJJ
of the sample duration.   The guiding principle in  choosing sample    h i>
is to  collect a  sample  over  the  longest  reasonable time period  ^ ten°
representative of the  condition to  be measured.    Longer samples  w   ^
to smooth out non-representative peaks  in the concentrations  wm
while sampling.

Detection  Limit.   The lower  detection  (LDL)  limit for  asbestos
is a function of  the sample volume  and the ratio of the  area  of
examined  to the effective filtration area.  The expression  is  as
              Af
     LDL =     r
            "V
where

    n  = the number of electron microscope grid openings examined

    Ag = the average area of an electron microscope grid opening

    Af = effective filtration area

    V  = sample volume

    The sample volume  should  be chosen to provide a LDL which 1s >
the  lowest  ambient concentration expected.   For rural ar635*
sources  of .asbestos  nearby,  concentrations  of  less  than  0*«*   a
            .                ,
structure/cm-3 are  often  found.   For  example,  in  order to achietply
        n limit  of 0.005 structures/cmj,  a sample of  approxlfflaie
liters  would  be  needed  if  10  average-sized  electron  micros
detection  limit  of 0.005  structures/cmj,  a sample  of approx
liters  would  be  needed   if   10  avera
openings are examined on a 25-mm filter,
                                    648
-------
                 Matter Loading  on  the  Filter.   The  presence  of  non-
       mater1al  on sample  filters 1s unavoidable.   In all  but  the  most
  er   cases,  the  asbestos  fraction of  the  total  mass  1s less  than
 be Jent.   The non-asbestos material  on the  filter can Interfere with the
^ivv  i  analysis  by obscuring asbestos structures.   If the loading is too
tie:ojj  (greater  than bout 25-50  yg/cnr),  analysis  by  the  usual  technique
^ f T i imP°ssi|:>le because a  uniform carbon coating cannot be formed  over
t!ie 11ter surface.  Alternative preparation techniques must  be  used, and
Coi)Ce J?esiJHs   of   the   subsequent   analyses    for   asbestos   structure
Crivu" trat1ons  are  not  comparable  to  directly-prepared samples.   In  an
the f{\ment  with even  moderate ambient parti cu late matter  concentrations,
**1eri  r  may become  overloaded  quite  easily.   For  instance,   if the
"•out \ particulate concentration 1s 30 yg/m  ,  sample  volumes  greater than
    J«200  liters  may  cause  filter  overloading.

         Analytical  Precision.   Because  airborne  asbestos measurement
      a  Count1ng procedure,  the  precision  of  the  method  is  inversely
      to the number  of asbestos  structures counted.   Polsson statistics
       the confidence Interval  for  the  estimate  of  the average  fiber
      On  the  filter,  assuming  the  filter  is  evenly  loaded 1n  a  gross
       The 95  percent  confidence  interval  for  the  estimated   asbestos
   en e  fading  on the filter is about + 100 percent when five structures
PerCe°Ur|ted.   At  a structure count  of  50, the interval   is about  +  30
hd Q!"   Normally, laboratories  will stop counting at 100  structures  or 10
Of ^nlngs,  which over comes  first.  Therefore, the best  one  can  hope
1 Unn  95 Percent confidence Interval of about + 20 percent.  However, 1t
        able  to  load the  f1lter   so  heavily  with  asbestos  that the
        *00  asbestos  structures  are observed  within  the  first  two  or
         °Pen1ngs.   In such  cases,  the loading for  the entire  filter is
        from  an  even  smaller than usual fraction of the total filter area
      1/50,000,  1f only one grid opening  is  counted.
         1  illustrates the  theoretical  precision with  which  the mean
     asbestos structure  loading Is estimated  as a function  of the mean
         Zt  ma^ be  seen  that  the  optimum  theoretical precision  may be
       at a filter  loading  of  10 fibers per grid  opening (about 479,000
       s  on  a 25-mm  filter)  because the area  scanned  and the  structures
       re  maximized.   The  right  side of  the curve  was  estimated by
       that  each  grid  opening  is  an  independent sample  and  that the
  9s e counts are  normally  distributed.   The Increase  in  the width of
 ty-ib Percent  confidence  Interval as  fewer  grid openings  are counted 1s
     f  to   tne   loss  of  degrees  of  freedom  in  the  student's  t-
     t1on.  This effect is not  strong,  however.
     ^rnber of  random samples, n, from a normally-distributed population
      "Jird  deviation a which  are required  to  estimate the  mean of the
        such  that  the probability is 1-a that the estimated mean differs
           mean by no more than d,  is given  by:
                                   649
-------
              Z2      a2

         "'-^-
                                                                        thfi
     In  this equation  z,   ,„ is the  standard  normal  variable for wn]c!lorne
 (one-sided)   probabilit/~°ar  exceedance  is 1 - a/2 .    Because  airj  ^
 asbestos concentration distributions are often log-normally distrlbuteoi  n
 is  appropriate to  use the  geometric standard deviation  of  the popu'° ^
 and  the difference between  the  mean  of  log-transformed sample data ana
 log-transformed population.

     If   it  is desired  to determine  whether the  (geometric)  means of  cj
 sampled populations  are significantly  different, the number  of  "'e('u
 samples is:


             <2iw2  + wL
        „ _
                   d2
 In this case,


              ima - mbl
        d =
where m& and mb are the true means of the  log-transformed  populations*   Of
a. and OB are  the  respective  standard  deviations.    The  probabi'1"
detecting a difference at the stated confidence  level  is 1-e.

Sampling Equipment
                                                                       four
    The  choice of sampling  equipment will  be  primarily  dependent  on   flf
factors:   the required  sample volume  and flow  rate,  the  availabi'   the
electrical power,  the practicality of manually  activating and stopP1
sampler, exposure to  the elements, and security.
                                                                       eji$
    If the samplers will  be  manually operated and 110-volt electrical P ^
is available,  a  carbon vane pump  with  a critical crifice will  serV!ii   1
well to provide constant  flow  rates  up to 20 1pm.  If the samplers *n
be  attended,   they  must  be capable of  starting  and  stopping
automatically  and accurately  measuring  elapsed sampling times
manufactures offer such samplers.

    In many  cases, outdoor  sampling locations  do  not  have  Il0
most available.   In  these cases, the investigator must  use  either
powered  pumps  or  provide  power  for   110-volt  pumps  with  a    te
generator.  However,  portable  generators are not acceptable for unaegni
sampling  because  they  generally  have  small  fuel   capacities  and  s
always stall  the moment no one is around.
                                    650
-------
   r"6 use of rechargeable,  battery-powered  pumps  with  automatic  timers  is
    convenient  for some  applications,  but  these  pumps are generally not
0,8, e  Qf providing  large  sample  volumes  or high  flow  rates.    With  a
   "to pore size 25-rnm cellulose ester-filter,  these  pumps  generally  cannot
    at flow rates  above  about  3  1pm.  The  rechargeable  batteries  will  last
    about  3  to 4  hours  with this  load,  producing sample volumes of  less
    ^5p liters.  The sample  volume  can  be  increased  substantially by using
         alkaline   batteries  with  these   pumps,   but   this   requires
     ation of  the pump  wiring to  take  the  internal batteries out  of the
          The   internal  batteries  are  not  compatible  with  an   alkaline
    •y.

   Bother option  for  attended pump  operation with  no 110-volt  power  is
   use of  a  iz-volt  car  or marine  battery  with   a  heavy-duty 12-volt
  U'  TMs setup will yield  flow  rates  of  nearly 10  1pm for 5 or 6  hours,
    Cln9 sample volumes of more than 3000 liters.

        PROBLEMS AND COMPLICATING FACTORS

       nt of Low Asbestos Concentrations in  Areas  with  High TSP
      problems  associated  with high  total  particulate  loadings  and  low
       ladings  on  sample filters  were  discussed above.   While  each of
%>on,p1rob1ems  singly  is  manageable, the  combination of  the two  can be
        troublesome.
     rural  areas which  have no  large asbestos  sources  nearby, ambient
Set     concentrations    may    be    considerably    less   than   0.001
ty ^**e$/cmj.   However,  these  same rural  areas  may have relatively high
Nbi   entrat1ons»  due  to  manufacturing  and   combustion  sources  or
Sup°Wn dust.   Residential  wood combustion can lead to TSP concentrations
       y Acceding 150 yg/mj in some  rural areas, and windblown dust can
       much  higher concentrations.    Under  such  conditions,  successful
        monitoring   may   be   impossible  with   the   usual   analytical
       s.   In order  to achieve a LOL  of  0.0005  structures/cm^ using the
    Ldboratory counting  protocol,  a sample  volume  of over 10,000 litecs
       ^quired.   However,  at  an  average TSP concentration of 100 ug/nr,
    . overloading  can  occur at  sample volumes greater  than  1000 liters.
P4ti0 *»  the standard technique  will  no± work in  such  cases.   Anytime the
     °f  TSP  concentration  (in ug/mj)   to   asbestos  concentration  in
            )  exceeds  about 30,000,   analysis  by  usual   technique  will
  i1h * fa11* no matter  what sample volume  is  used.   In  order to obtain
   n9nn  results,  the  ratio  should  be  at  least  300,000:1.

JH^J. Alternative analysis technique which would be  appropriate  for such
Ntims  wa* described by Yamate,  et  al.  (1984).   The technique involves
 1$ rPerature  ashing  to remove organic matter  from the sample.  However,
        ique also requires that the sample  be  resuspended in water using
         and refiltered  before  counting  under the TEM.   Experience has
         that   results  of  samples  prepared  in  this  manner  are  not
                                  651
-------
 comparable  to those  prepared  in  the usual  manner due  to the
 bundl'es  and  clusters.   Therefore,  all samples should be prepared
 if meaningful comparisons  are to be made.

    Another  possible  solution  would be  to  arrange  for the  1 abora
 examine  a  much  larger area of  the  filter.   In this manner,  smaller
 volumes  could be used  in order  to achieve  the desired  LDL.

                                                                      -—
Choosing Sample Volumes without Prior Knowledge of Asbestos ConcejitraJ

    A  common  problem in outdoor asbestos  sampling  is the lack of fi
knowledge  of  the  magnitude  of  asbestos  concentrations to  be  i"e*
Outdoor  asbestos  concentrations have  been measured  in  very few are
the investigator  can only guess at the  ambient concentrations  in  tne
area.

    Fortunately,  TSP  concentration  data are available for many ar.ea**e  ^
areas   that  have   no  data,   an   educated,   conservative  estima te j
sufficient.   The  estimated  TSP concentration  can  be used  to ca ^ices °
maximum sample  volume that will avoid overloading.   If  no known sou ^\t
asbestos air  emissions  are in or near  the study area, the use of
volume close to the maximum will probably work well.                     ,
                                                                      rA$
    A  better  solution yet, in such cases,  is to  operate several co1    of
samplers  at  different flow  rates  to  collect  samples over  a  r* ?  jr
volumes.  The  most appropriate sample can be chosen at  the
a long-term monitoring program  is  planned,  this approach would be
only for the first few sampling days.

Choosing the Number of Samples without  Prior Knowledge of  the_Var—-^
Ambient Asbestos Concentrations

                                                                    "*A $
    As described  in  a preceding section, the  number  of  samples    u^   *
estimate means  or detect prescribed differences  can  be  estimated,      a
the  standard  deviations  of  the  sampled   populations.    In  9en
investigator will  not have  the  benefit of  any data  regarding  '
of airborne asbestos  concentration  in  the project area.   However,
often data which can serve as a surrogate for  this information.

    If the primary  source of asbestos  in the  study area is w^nclb!5g uSe^i
available  TSP   concentration data  for  the area can provide  som  ^Iflfj
information.   Although  the asbestos concentrations may  not be  PJ^y to
to TSP,  the  variances of  the  log-transformed  populations  are  11k^se
similar.   However, because asbestos  measurements  are  much less  Pr?*ofis
TSP measurements, the variance of the  measured asbestos  concentrat   ^
likely   be   higher.     If  the   geometric   standard   deviation
concentrations   during  meteorological conditions  similar to  those
during asbestos  sampling is 1.0, it would  be  prudent  to  predict ^
geometric mean  of the  measured  asbestos concentrations  will  be
2.0.
                                   652
-------
     other  known  sources of asbestos  emissions  are present in  the  study
  ^c  reasonable  approach  would  be to  estimate  the  emissions and  use
 - ftvp  a^r ^aUty modeling.   It  is not  important  to accurately estimate
'" the  "e  emiss1on rate for these purposes, only the  relative  variation
fOi-tu4rate and  the variation  in  meteorological  conditions are  important
   nis approach.
  If
i^arrf  ^ata at a11 dre  available, a very crude  estimate of the  geometric
Sifist  deviation  can be  made  as  follows:   guess what the  highest  and
'Ojaru C°nce1vable  concentrations might  be;  divide the  difference of  the
p    nms of these two values by 6, then add 1.5 for analytical  variance.

            Large  Sample  Volumes  with  Self-Activated.  Battery-Powered


     an  investigator   finds   it  necessary  to  collect  many  samples
    "eousiy and in remote  locations,  the use  of self-activated, battery-
   j..,- -~r- will certainly be necessary.  However, due  to  the  low maximum
1?" Can   Wn1cn  can  be achieved by  these  personal  sampling pumps  and  the
!CfiSsav?Clty  of  the  Eternal,  rechargeable   batteries,  it will  often  be
  kcah  to e^ther attach several pumps  to one  filter, or use  an  external,
    Paclty battery to collect a sample over a long time interval.
              of   outdoor  ambient   airborne   asbestos  concentrations
       some  new  challenges  to  the  investigator.     Outdoor  sampling
    'J are  different  from indoor  programs in  a number of  ways, and  no
    yuidance 1s available.  In  designing  such  a program, an  investigator
     consider  all   available data  to  help  choose the  most  appropriate
    ^J1   method,   number   of    samples,   sample  volumes   and   sampling
     *•     Through careful  experimental   design,  an   investigator  can
(.  - 'ncrease the  probability  of obtaining useful  information.
L
> «.
^i§Cti
(Pi
1984.
i Micros
Methodoloqy for the Measurement of Airborne Asbestos by
•copy. EPA Quality Assurance Division.
L
\
      k
      6'   Measuring Airborne Asbestos Following an Abatement Action.  EPA
      -85-049.
  W  G'J.   and  A.P.   Rood,   1983.     "Membrane-Filter  Direct-Transfer
  W? ?Ue  for  tne  Analysis   of   Asbestos   Fiber  or  Other   Inorganic
  I?, :c'es  by Transmission  Electron Microscopy"  Environ.  Sci. Technol.

 \      "    '
   '  986»  NIOSH Manual  of Analytical methods, Method  7402.
                                  653
-------
    Effect of Filter Loading on  Precision



£
Si
•D
§
"6
V
c
ID
u
c
2
c
0
0
K
in
o>


JOVJ -
360 -
340 -
320 -
300 -
280 -
260 -
240 -
220 -
200 -
180 -
160 -
140 -
120 -
100 -
80 -
60 -
40 -











I
1 ^^^
\^____ _______ 	 	

20 ~\ | ] j 	 1 	 1 	 1 	 , 	 , 	 1 	 , 	 1
024 6 8 10 12
                           (Thousands)
                Asbestos Loading (Structures/cm^S)
Figure 1   The effect of filter  asbestos loading on precisi0
                        654
-------
 r >•
Ij  eParation  of  Summa  Canister Performance Samples and Their
 e NJ 08837
      Pritchett
     Environmental  Response  Team
       08837
 As
  * Part of  the  QA/QC  plan  for  the indoor air portion of the
  anal Emergency Declaration Area Habitability Study, Summa
  er performance evaluation  samples were prepared and
  *ed by NSI-ES according  to specifications defined by the
  *PA Environmental  Response  Team(ERT).  These samples were
      zed in  the field  by  the  TAGA 6000E MS/MS using
  u*"es developed by the ERT  and Roy F.  Weston. Preparation
     sis procedures will be  discussed along with various
     which  had  to be  overcome  initially. TAGA results for
  analyses  will then  be presented along with their deviations
    ected results.  Finally,  these results will be compared
 0  the data  quality  objectives and summarized with the
  riate conclusions being  drawn.

                     INTRODUCTION

 gj*-ES was  directed1  by the  Quality Assurance Division of
 a SL/RTP to provide QA  support to EPA's Love Canal Emergency
 4 ation Area  Habitability  Study at Niagara Falls, NY.
 n,lftajor portion of  this support was the preparation of blind
 1 16L summa  canisters  containing the two selected Love
  fndicator Chemicals(LCIC's)  chlorobenzene and either
    toluene or  o-chlorotoluene. These canisters, prepared at
  ^50 PPM ]_eve;i_ Were  used  to check the  performance of the
       Mass Spectrometer/Mass  Spectrometer instrument used
    study.
     report places  emphasis  on the cleaning of canisters
 ".to Preparation, preparation,  NSI analysis by GC/FID, the
 (uQalysis results,  and  comparison of the TAGA results with
 p\ °bjectives outlined  in  the  Quality Assurance Project Plan
   *°r the  study.

                    EXPERIMENTAL
                           655
-------
     Typically,  the  NSI  Volatile Organic Standards Laboratory   g
 prepares standards  in  the  0.5-5.0 PPB range and it is impera
 that canisters  be thoroughly  cleaned before using them. Sue"  jj
 cleaning is no  less  important for canisters which are to cotl  e
 more highly concentrated  species, such as in this study. ^ ° g<
 1 shows a schematic  drawing of  the canister cleaning apPara ,g-'
 It consists of  a vacuum pump  with an ultimate vacuum of 1 x ,,j
 Torr, a V coiled copper  tubing trap, a Dewar flask for li(Jyf gjt
 N2 , a thermocouple  vacuum gauge with a range of 1-1000 && °^
 a bubble flowmeter,  an  oven suitable for heating 2 6L canis
 simultaneously, 1 3-way valve,  3 2-way valves,  and copper
 connective tubing.
     Prior to canister  cleaning,  the  copper trap is cleaned   ^
 remove acculated contaminants and H20. With valves B and D
 the trap is purged with high  purity  He or N2 while heating
 with a heat gun for  about  lOmin.  After the coil has
 valves B and D  are closed, the  Dewar is replaced and
 filled. Valves  A and C  also remain closed when  the vacuum
 is first turned on.  Valve  A is  then  opened, followed a^teirirt/J*'
 minutes by C. When the  vacuum gauge  indicates a vacuum °^/S")
 the 3-way valve, D,  is  opened to  the canisters. The Nupro^.^'
 canister valves remain  closed until  the gauge again reads
 One of the canister  valves is then opened and that catilster f
 evacuated to 500 Aim. Its'  valve  is then closed  while the ° efleum .
     In order to determine  the efficiency of cleaning,  eva<  (jj
 canisters are repressurized with  zero air,  dry  or humidif* e
at the septum fitting. After  injection,  any  remaining lifl
the needle is again drawn  into  the barrel  and mesured,  s°
                                656
-------
  Ift a
  Pte88,^Ual volume of  liquid  injected is known. The absolute
    "~^at the  time  of  injection is also noted.Mild heat is also
       in the injection area  to  facilitate evaporation. After
  -=8s   tne canister  valve  is  opened fully and the desired final
  >1<1 t,re reached in 20-30min.  The canister valve is then closed
  f«checf canister allowed to  sit  for about Ihr and the pressure
  ^ Chi     This  pressure is  then  noted and the theoretical PPM
     °robenzene and  chlorotoluene determined.

               Analysis-Calibration

  *f>Ur  itration curves were  prepared using diffusion tubes(see
  l^cti     containing  chlorobenzene and p-chlorotoluene as the
   fUs?ning elements  of  a diffusion chamber calibration system.
  '"d c  °n of material through  the neck of the tube is predictable
  Jvej.  H be measured very .precisely.  By measuring weight loss
  'Ud 9tl extended period  of time,  during which the tube is being
  !a8{8at a constant temperature with a  constant flow of carrier
  r th    as N2^  across the tube(see Figure 4),  an accurate figure
  ^•.  e quantity of diffused material  in the  gas stream can  be
         This in turn develops a  very precise primary standard
        traceable to NBS. Using  basic diffusion theoryS, the
              ,  weight  loss, and  a programmable calculator  the
   jj»-Lon rate for each  component  may  be determined  in ng/min.
  ambfi       point calibration curves are prepared by  trapping
   the* stream samples cryogenically while holding  the flow  rate
  l^iy  traP constant and varying  the trapping  times,  then
 5*Hll   g them uslng a Tracor 550  GC/FID.  An  OV-17  glass
 "8ed. Aty  Column  50m x O.SmmID with  a  2.5  urn  film thickness was
  D6t  dditional  GC  parameters were  as  follows:
  C01,ct°r  Temperature               250*C
        Flow                         5cc/min
       •  Up Gas  Flow                 30cc/min
      °cusing                       Imin before  end of  trapping
      Program                       t.£ = 30aC
                                   hold  1 minute
                                    close  oven door
                                   program at 5°C/min
                                    tf = 130°C
    • !!_  ana^yses have  been completed  ng  are determined  and  in
         converted  to ju.% before plotting with the corresponding
       »r  the entire  period of interest,  chlorobenzene gave  a
       response  for Aig  vs.  area. p-Chlortoluene showed  some
    ;ring  at  higher concentration levels. However, on a  curve
    *?e basis, it was found  that  23 of 25 correlation
     ;ients were greater than  0.995 with 8 values of exactly
  ^  ^ single low  value of  0.984 was  found for an unused
  8r6ation eurve' F°r p-chlorotoluene  14 of 16 values were 0.995
 uflresat:er with 8 values  exactly  1.000. The lowest value of 0.987
,    Ponded to the  same  run  as  chlorobenzene.

  ^^-~-£§JLi ster Analysis - Can is t e r s

^ te Ug tne  Same GC/FID conditions as for chamber analyses,
|>? Of Plicate  75mL samples from  each canister  were analyzed and
i,^ t,^cnl°robenzene and chlorotoluene dtermined for each run.
t  a6 calculated for  the  individual  runs  and the means of
           e values used for  reporting  overall canister
     rations. Figure 5 illustrates  a  typical  chromatogram

                             657
-------
with the appropriateyug and PPM.

                   RESULTS AND DISCUSSION

    The QAPP for the Love Canal Habitability  Study  lists
QA objectives but this report deals only with  the two
canister related areas, TAGA 6000E accuracy and  precision>geg
    TAGA accuracy was to have been determined  in all 4 p"
the study using blind canisters supplied by NSI-ES. How*
during the early stages of the study the canisters  were
as true performance evaluation samples because NSI  analys^£
procedures were still in the development stages  and a
had not been determined for acceptance or  rejection of ca
A major problem that NSI had to deal with  in  analyzing t  Q t^
canisters was the concentration level that was necessary    (
the TAGA could efficiently analyze the field  diluted samp^
A gas dilution system was unavailable and  in  order  to avo  -•*$>• '
overloading the GC column and detector a relatively small
75mL (5cc/min canister flow, 1.5min collection time), waS
analyzed. At these small levels errors are much  more sis*1
There were also problems with build up and/or  hold  up of   . (be
compounds in the lines and valves between  the  canisters a
cryogenic trapping system for the GC. These problems werehe
compensated for by allowing the sample to  purge  through    i
entire system for a few minutes before sample  co llectioti   ^ f
purging the system between canisters with  zero air  or he
2 hours with the GC oven at 200°C. The purge  stream was 3 ^ e^
analyzed before proceeding to another canister.  Also, a* ^
day's analyses, the system was purged overnight  with the
oven at 200°C. Thus, during these developmental  stages, c jjd3*
instrument accuracy was primarily measured using Scott s    J
cylinders of the type used for TAGA calibration. The ana
cylinder was never the same cylinder which was used f°r
applicable calibration. Starting in phase  2 the  16L
were also used to determine the overall accuracy of the
Finally, in phases 3 and 4 the 6L canisters were used to
the relative accuracy of the instrument.
    The accuracy criteria to be met by the TAGA  was that ^gg
magnitude of the error in its' analysis values be 25% °r
Relative accuracy as measured by the Scott cylinders neVfouf
exceeded an absolute value of 25% on any day  during the    -
mobilizations. The largest magnitudes of the  relative er   pZ
measured by this analysis were 23.1% and 23.0% for  chlor°
and chlorotoluene, respectively. During phases 3 and 4,  ^ 3p
relative error measured by the 6L canisters never exceed   ^j,
absolute value of 25%for either compound.  Overall,  for c  ,g(j
analyzed by both NSI and the TAGA the 25%  limit  was    °e
times for chlorobenzene and 4 times for chlorotoluene.
Investigation showed that the first two of these
differences exceeding 25% because of a problem in the
delivery system which was later corrected. Table 1
NSI theoretical PPM chlorobenzene and chlorotoluene, NSJ- ^g
analysis values, TAGA analysis values, and the % differe ^
from the theoretical values. The magnitude of  the relati
error for the compounds as measured by the 16L canister
analyses only exceeded the 25% criteria once  for chloroto
only during phases 2-4. However, on that day  the magnitu ^
the relative errors for chlorotoluene as measured by the
canisters and the Scott cylinder were 8.8% and 8.0%,

                             658
-------
d^^ively. Figure 6  illustrates  graphically the TAGA %
aga.erences from the theoretical  concentrations when plotted
t0^ nst analysis order.  In  general,  chlorobenzene and chloro-
V   fine follow the same  pattern showing  a positive bias. Mean
clUoe8 °f + lt7 -  14'A for  chlorotoluene and +2.5 + 19.2 for
HSj r°toluene confirm  this.  Figure 7  shows a similar plot for
atij atlalyses. For  the  most  part chlorobenzene shows a + bias
the C^°rotoluene  compliments  it  with a  - bias that is almost
the B>irror image °f the  chlorobenzene curve. Using this data
chj Bleans 8.8 + 14.6 for chlorobenzene and -8.4 + 12.2 for
detto toluene were determined  and  2  x SD for each set up to
fot rmine data acceptability.  These  values are 29.2 and 24.4
So Qclllorobenzene  and  chlorotoluene,  respectively. Out of the
lhe c s° canisters prepared  only  three,which were not used In
   ~;eld, exceeded these limits.
   /AGA precision was  determined by periodic cylinder/canister
        Again the criteria to be met was 25%. Day-to-day
     '•on of the TAGA  was actually determined from 2 sets of
J6^icate analyses. First,  as  per  the QAPP,  the precision was
 ^^itied from the daily analysis of the 16L summa polished
         The maximum relative standard deviations measured in
    analyses throughout the  study  for  chlorobenzene and
   'toluene were 15.6% and 17.9%,  respectively. Because of the
dfivi  number of analyses for  each sample, the relative standard
'tiaiat*-°ns were also calculated for  Scott standard cylinder
t!io8y8es* The maximum  relative standard  deviations measured in
3,5 e analyses for chlorotoluene  and  chlorobenzene were 5.4 and
*H ' tespectively. Both  sets of data  demonstrated that the TAGA
   ^sis met the required data quality objective of 25%.
^ n summary, it is  clear  that  as   a whole the TAGA 6000E
   Spectrometer/Mass Spectrometer  analyses met both the QA
         for precision and accuracy. In those instances where
       not occur the problem was quickly located and corrected,
                ACKNOWLEDGEMENTS

The authors would like to thank B.J. Carpenter,  Shirley
>  Annette King, and Karen Ol
lleir contributions to study.

                   REFERENCES
       autors wou       e  to    an    ..           ,
fot y> Annette King, and Karen  Oliver  at  Northrop Services,  Inc.
        support provided  under  contract  number 68-02-4444
   W1th the US EPA.

   J- F. Cuthrell and W.L.  Zielinski,  "  A Gravimetric
   Technique for the Preparation  of  Accurate Trace Organic
       Standards", APCA meeting,  Boulder,  Colorado,  October
    .
    A1tshuller and Cohen, Anal.  Chem,  802,  1960
                            659
-------
         Figure 1. Sch«»«tlc of Ctnliter Cleaning  App«r«tu«
              ZEIO All
             ^BALANCE GAS
                                                  FEHRVALT ABSOLUTE
                                                             CADGE
                                                 •INJECTION  SEP TDM
                                                  4-CAKISTEl
Tlfur* 2.  Apptntu*  for  C*ol*C*r  trtpttmtlon Uiing a Haitcr Solution
                                   660
-------
                                                                 Olffuiion
                                                                 Tub*
                    Figure 3. Diffmlon Tub*
            CARRIER OAS OUT


                    I
O-RINO
                                 STAINLESS
                                   STEEL
                                 CYLINDER
                                    FOR
                                  SPACER
                                                                ISOLATION
                                                                 TUBES
           CARRIER OAS IN
       Figure  4.  Diffusion  Chaabcr Calibration Syit*B
                                 661
-------

                                                             t.«0.
 CIT  ire 0.40
 mo    10.0
 «TI It   S
 «I fGIL  *
 Itf lint  0.3J
 Mil IIJICT  WO
0.00 flf/IIT  -1
Lit flf/HT  J
1.10 ftf/III  )
t.IO »L»/m  1
1.50 fLf/IIT -»
                                    M
       CILOIOlEHtll   |.01

       CKOlOTOLgill   It.10
  41U     JSS.      tin

*«.illO   1.J7JJ    14.J

tl.1100   l.lOil    JJ.J
                           T.bl.  1
CANISTEK
ID
LC-0027
LC-0037
LC-0031
PEB-2
LC-OU07
LC-0070
LC-0040
LC-0009
LC-0012
LC-0003
LC-0063
LC-0047
LC-0039
VEB-1A
PEB-2A
Pit- IB
LC-0013
LC-0025
LC-0026
LC-OOS2
LC-0069
LC-0072
LC-0066
LC-0013
LC-0030
LC-0033
LC-0049
PEB-1B
PEB-2B
LC-0040
LC-0003
LC-0056
LC-0012
LC-0020
LC-0039
PEI-1C
MSI THEO
CBEHZ
38.6
32.23
26.37
34.4
23.45
32.52
47.8
47.6
31,44
28.78
42.13
28.27
28.81
36,6
37.55
40.63
37.33
34.67
41.07
33,60
37.33
31.47
37.33
37.33
37.33
37.88
30.40
40.63
37.10
37.30
28,24
35.17
33.57
43.71
38.70
37.0
CONCS
CIOLU
33.2
27.74
22.66
29.6
20.16
27.94
40.77
36.64
27.02
24.75
36.52
24.5
24.96
31.49
32.28
34.89
32.36
29.78
35.27
29.12
32.06
27.03
32.66
32.36
32.36
32.88
26.33
34.89
31.87
32.36
24.27
30.23
28.86
37.88
33.26
31.78
MSI AMALD
CBINZdDlM
39.2(41,6)
37.0(414.8)
25.6(-3.0)
40.3(417.2)
33.9(444.3)
31 .5(-3. 1}
43.5(-9.0)
47.4(-0.4)
36.7(416.9)
36.0(425.0)
40.4(-4.0)
34.0(420.1)
26.2(-9.0)
47.0(428.4)
41.2(49.6)
42.1(43.6)
39.4(45.5)
35.7(43.0)
32.3(-21.4)
38.2(413.7)
39.1(44.7)
34.2(48.7)
34.3(-7.6)
39 .4 1*5.5)
34.4{-7.4)
39.2(43.4)
32.0(45.3)
42.1(43.6)
37.1(0.0)
41.7(4-11.8)
34.5(422.1)
41.5(418.0)
36.1(47,3)
40.6(-7.1)
36.5(-5.7)
38.4(43.8)
CONCS
CTOLU(XDlf)
32.0(-3.6)
18.4(-33.6)
21.8(-4.0)
25.5(-13.9)
1S.9(-21.3)
27.9(0)
33.9C-16.9)
23.8(-35.0)
26.5(-1.9)
24.6(-0.8)
25.4(-30.4)
23.K-5.7)
25.5(42.0)
33.3(45.4)
30.6(-5.3)
34.3(-1.7)
26.8(-17.2)
2B.9(-3.0)
32.3(-8.4)
31.3(47.5)
33.4(44.2)
26.8(-0.9)
30.K-7.8)
26.8(-17.2)
29.0(-10.4)
27.3(-17.0)
23.K-12.3)
34.3(-1.7)
30.2(-5.2)
2B.7(-11.3)
23.41-3.6)
29.71-1.8)
29.6(42.6)
33.8(-10.8)
33.3(40.1)
30.4(-4.3)
TACA EISULTS ,j
CBENZ(ZDlf) CTOLJLii^1-^'
26.K-32.4)
22.4(-30.5)
25.6(-3.0)
31.7(-7.8)
27.4(416.8)
32.2(-1.0)
48.9(42.3)
43.8(-7.9)
33.3(45.9)
39.6(437.6)
40.6(0.6)
21.6(-23.6)
31.6(49.7)
37.9(43.6)
36.3(-3.3)
46.1(413.5)
42.5(413.8)
34 . 3(-l . 1)
42.7(44.0)
42.0(425.0)
41.0(49.8)
32.5(43.3)
41.0(49.8)
37.4(40.2)
37.9(41.5)
36.9(-2.6)
32.6(47.2)
40.21-1.0)
40.03(48.0)
42.9(415.0)
30.0(46.2)
39.9(413.8)
37.5(411.7)
15.01-19.9)
30.2(-22.0)
38.23(43.4)
"•s!:iJ:ll
** • i a)
nft I -4 »•"
* 1 li 7)
36,9(4»*'"
23.*(+l6:)!>
28. 7(*'' ii
34.8(-«*iJ'
39. K**'i
29 • * I* I »)
3i»2(** %
36. 6(*^! \\
* . * * 8)
30.9(+zJj,
30.8("*'.j
31 ,7("'*
40.7(41^*'
31.0(-*'J
29 ,5(*°' jj)
38 .8(4 !»• ,j
« .< i|
36.6(+1*j,
29.4(+(<
36.9<+lS'0)
27, 5^1)
34.1<+*;j,

29 . 1 (*^*?
36.2^*^!* A
"•"ft? i>
"•'Jlli »
27.1J+ J ,)
"••{Hi 1)
34.0(4lJ'Jj
30.9(-|^;5)
* Z/ tja j)
34.42(+»'
                              662
-------
6,TRGfl  % DIFFERENCE BY RNflLYSIS ORDER
            10          20          30
             flNHLYSIS ORDER OF CflNISTERS
   NSI  % DIFFERENCE by RNflLYSIS ORDER
                             j.i	11 tin r	'
                 20      30      40      S3
              RNflLYSIS ORDER OF CftNISTERS
                   663
-------
  THE IMPACT  OF RESIDENTIAL  WOOD  COMBUSTION  ON  AMBIENT WINTERTIME_C
  MONOXIDE CONCENTRATIONS IN RESIDENTIAL AREAS  IN SIX NORTHWESTERN
 James E. Houck, Carl A.  Simons
 and Lyle C. Pritchett
 OMNI Environmental Services, Inc.
 10950 Southwest Fifth Street, Suite 160
 Beaverton, Oregon  97005

 Gerry C. Snow
 OMNI Environmental Services, Inc.
 Post Office Box 14001
 Research Triangle Park, North Carolina  27709


      During the  winter,  elevated  carbon  monoxide  levels  in  residen
 areas  in most  Northwestern  cities  originate from  the combined  im
 vehicular  exhaust  and  residential  wood  combustion  (RWC).   The  r
 impacts  of the two  sources have been traditionally  difficult  to se
 In the  study  reported here, the  portion originating from RWC was cfllc
 for residential areas  in six cities where  woodburning is  recognizgd
 significant source of air  pollution:  Boise,  Idaho;  Eugene, Oregon;
 Montana;  Missoula,  Montana; Portland, Oregon; and Yakima,  Washington-
 use of fine particulate,  nephelometer and  total  carbon monoxide  recoi
 the relationship  between  carbon  monoxide  and  fine  particulate  eDl
 characteristic  of RWC  permitted RWC  carbon monoxide  impacts  to t>e
 mated.   Approximately  100,000  data  points  were utilized  in the c
 modeling.   In addition  to  the  quantitative calculations temporal P
 and meteorological  correlations  provided   supporting information
 magnitude of the RWC  carbon monoxide  contributions  that were calcul»
                                                                     to
     The average  percentage of  ambient carbon monoxide attributable   ^
 ranged from a  high  of  67  percent  (1.86  mg/m3)  of  the total ambie°^int«*
 centration  level  at  the  Boise,  Idaho  monitoring site during  col<*  tflt*J
 Sundays  and holidays  to a  low  of  13 percent  (0.28 mg/m3) of  tbe ettfe
ambient  concentration at the  Yakima, Washington monitoring site »v ^ $fi
across all  winter  days.   In addition  to assessing  the contribution ° tjj«
 to  atmospheric carbon monoxide levels,  the  approach presented here & ^ o'
potential to permit the  contribution of RWC  to  ambient  concentrate^ fl{J»
toxic air pollutants  such as PNA's  and aldehydes to be estimated bflt
short-term and on a long-term basis.
                                    664
-------
       air  quality impact of  residential  wood combustion  (RWC)  has  been
 tot " to  be   significant  in  many  locations,   and  the  impact of RWC  on
 L lculate concentrations  has been the subject of numerous  studies  in the
      t.    The  impact  of RWC  on ambient  carbon monoxide  (CO)  levels,
   |*eri has  been much less well studied.   On  a  mass basis,  RWC  produces
    '*imately  six to  ten times  the amount  of  CO as  it does  particulate
          The potential human  health impact due  to  RWC is aggravated  by
       s  that  the  majority  of RWC  emissions  occur  during a five-month
       IWC pollutants  enter the atmosphere at near roof level  in  residen-
 't»j areas,  and wintertime  temperature  inversions  cause  localized  air
    ation in many wood burning  areas.

 of c ^esidential wood  burning  and motor vehicles are the two major sources
 CO c  *n  residential areas.   The fundamental problem in assessing the RWC
 of retribution in residential areas is apportioning the  relative  fractions
    Derived from each of  these  two sources.  Unlike particulate material,
    has  identifiable chemical  and physical characteristics which  can  be
 ipp  to distinguish  specific  source  contributions,  other,  more  subtle
 V^  aches must  be   used  to apportion  the  sources of  CO.   Another  factor
 to ho complicates CO source apportionment is  that  short  time  periods  need
    examined.  The physiological effects of CO result from  short  exposure
      and the  corresponding  health-related primary national  CO  standards
    °r one-hour and eight-hour averaging periods.

 "W^6 results  of  a  study which  estimated wintertime CO  levels in  six
 *itie Vestern  cities  are  presented here.   Hourly CO data, hourly and daily
 tL  Particulate data  (<2.5p  particulate  sampler  data  and  nephelometer
 ((>(,. '• Meteorological data and temporal patterns of woodburning and traffic
 Iti r^ties were utilized  in estimating the RWC impact on ambient CO  levels
             areas.
tet, Tlle ratio  between CO  and fine  (<2.5/i)  particulate  emissions charac-
V!tiC of RWC and  the  relationship between the light scattering  (b-scat)
Sc    ristic of  wood smoke  particles and  the  corresponding particulate
Vsntration were the two key Parameters used for estimating RWC contribu-
I>Sft.  to  ambient  CO  concentrations.    The  mean  ratio between  CO  and  fine
^t   late emissions  (mass ratio)  for RWC  is  8±2.1-3   The  relationship
    11  particulate  mass  and b-scat  values for  wood smoke  is generally
    as between 1.5 x 105 pg/m2 and 2.0 x 105 /tg/m2.  The slope of m e a n
  ,,'  fine  particulate  concentrations  measured with  a GCA Corporation
  ^ °r Plotted versus mean hourly b-scat values was  1.46 x 105 /ig/m2 for
T^te dflta  sets  (R  =  '89)  obtained  from Missoula,  Montana,  during  the
* dj_ £• and the slope of daily fine particulate concentrations sampled with
'Hs  ot°mous sampler plotted  versus mean daily b-scat values for 123 data
J-4i ^ *  .93) obtained  from Portland,  Oregon,  also during the winter, was
   1 ^^ /ig/m2.   The  differences  in  slopes  between  those  measured in
       and Portland and the values generally cited for pure wood smoke is
     y due to the  facts  that the coarse (>2.5jO particles  present in the
       ce  contribute  to  the  observed  light   extinction   (lowering  the
    1  and  that  there   are  other  sources  of  particles  in  the  airshed
    8  those  originating  from  woodsmoke.   Winter   was defined  for  the
       of  this  study  as  November  1  through March  31.    Holidays  were
         as Thanksgiving, Christmas and New Year's Day.   Cold winter days


                                  665
-------
                                                                      c
  were defined  as  those days with  a  heating degree day  (HDD)  value  g
  than one-half the standard deviation above  the mean winter  HDD value-
  monitoring sites were selected for  their  residential character.
  monitoring sites  were:  (1) Mountain View Elementary School  - Boise,
  (2)  Amazon Park  -  Eugene,  Oregon; (3) Wrecking Yard  -  Libby,  Montana-
  Lions/Rose Park  -  Missoula,  Montana;  (5) The  Intersection  of 58t .
  Southeast  Lafayette - Portland,  Oregon;  and (6)  County Courthouse -  *fl
  Washington.   Only  time  periods when both  valid  mean  hourly CO  valu
  valid  mean  hourly  nephelometer  values  were  available  were  used
  study,  so  that the  estimated RWC CO component  could  be compared «l
  total  atmospheric CO  concentration.   Approximately 30,000  CO-b-scat
  were used  in  the  study.   The estimated RWC  CO component  was an uppef
  estimate,  as  it  was  inherently assumed  in  the  calculations that al   tc
  particles  in  the  airshed were  from  RWC  and the larger fine  partiCUjclllflte
  b-scat  ratio  value  (20  x  105  /xg/m2) was  used.   A  number  of  part* ^ 0{
  studies have  been conducted in the six cities and a reasonable  es  . - $
  the  percentage of  fine  particulate  concentration  due   to  RWC dur *
 winter would  be 602  to 902 of the total.  Both short  term diurnal
 hourly  CO  levels  for special  event days  and wintertime  average
 plots (by  the hour) are presented.  In addition to the direct  calc ul
 the  effect of  temperature (heating  degree days), vehicular  activ    jn
 woodstove  activity   on  total  ambient  CO  concentrations   was  use   ^
 illustrating  the  relative CO impacts  of RWC and  vehicluar  exhaust
 residential areas.
of6"-
 Results

      Residential wood combustion was shown to be a significant      h
 at  the  residential  monitoring  sites  during  the  winter with  ve  t C"
 exhaust  apparently  the major  source.   The  maximum percent  of  *D1^ acf1!
 attributed to RWC at  the  six sites ranged from  131  to 392 averaged   $
 all winter days, from 152 to 532 averaged across cold winter days oO 7' rt»
 from 162 to  602  averaged  across  Sundays and holidays  only (Table *^j>t<
 highest percentage of  the total  ambient CO derived from RWC was calc ^
 for the  Boise  monitoring site, which has  been  designated  by    S Of
 Department of Health and  Welfare as a "wood smoke" site.   A
 of the ambient CO on cold Sundays  and  holidays  was assigned to «•-  .0 •• (
 site.   Not  only was the  estimated percent  of  ambient CO  assigned    J."
 higher on cold  days, but  the total ambient  concentrations  ranged £ ^,
 to 482 higher for  all  sites on cold days  as well (Table I).   **°rt,
 higher vehicular CO emission factor, and a greater probability of
 ture  inversions  can be expected  on cold  days.   The impact  of  c ^
 weekend and holiday  traffic in residential  areas  can also be  seen.,j
 data.   The  decrease in weekend and holiday ambient CO levels  rang ^
 202  to 302 at the  six  sites (Table I).  Limited data  for  Northwest
 show  that weekend traffic  counts in residential areas  are  152  to  2
 than weekday counts.

     Diurnal plots  of  total atmospheric  CO  and  the  corresponding ^
 limit  RWC CO values  are illustrated in Figures 1  through 4.
 term special event  days and  long-term  (winter-long  or multiple
 sets can  be  reviewed by utilizing  the real-time  output provided by
meters.   Figures 1  and 2 are  diurnal  plots of mean  hourly data
winter  of monitoring in Eugene.   Figure 1 illustrates the  diurn**
averaged  across  all winter  days.   Figure  2  is  for all winter  Sun
holidays.   Rush-hour traffic is generally  considered  as occurring
0700 to 0900 and  1700 to 1800.  Peak wood burning hours are  between

                                   666
-------
 VII
fte  and  between  1700  to  1900.    The  last  wood  load  of  the  day  most
'ion  ntiy occurs  between  2200 and  2300.   Ground-based temperature inver-
ict,  generally start forming  in  the evening.   The combined effects of the
|hollijty  patterns  and meteorology  can be  seen in  Figures 1  and 2.   It
'* (U  ^6  emPnasized that poorer dispersion causes  a build-up of pollutants
   B^t,  and the plots are of atmospheric concentrations not of emissions.

     rllaPs  the most valuable application  of  the  methodology developed in
         is  tne  examination of short-term events.   Figures  3  and  4 are
      Plots of two 24-hour periods  at the Boise  monitoring site.  Figure
    strates a cold day when  vehicular  traffic was at a  minimum, and RWC
>ehic  a near maximum (Christmas); and Figure  4 illustrates a mild day when
•*• Ca     traffic was high  (Monday), and  RWC was  probably  relatively  low.
   n be seen  the total atmospheric CO was  on the average higher on the
     daX,  and the  bimodal  traffic-related impact was more  pronounced on
   atter day.   The RWC-derived CO was much higher  throughout  the  cold
  CO   ,8  day-   The  ability  to apportion  the  sources of CO during periods
»t   Eolations make  the  methodology presented here particularly useful.
\ *an>ple,  during  an  eight-hour CO violation period  (1900,  11/10/82 to
\'  ll/ll/81)  at  the Portland monitoring  site, it was  calculated that, at
   l|nmn, 24 Z  of the CO was  derived  from  RWC.
  I
C(iri*att
V/116 uPper-limit  of  atmospheric CO produced  by  RWC was determined  for
%(, ential monitoring sites  in six Northwestern cities during  the winter-
\^  The RWC  derived  CO was  found  to be  significant,  albeit vehicular
1lsn8t was  the  major source  of atmospheric CO.  The  real-time nature  of
Sae?helometer data allowed diurnal trends to be studied.  In  addition  to
 je ih8  the  contribution of  RWC  to  atmospheric CO  levels,  the approach
     d here has  the  potential to permit  the  contribution of RWC to  the
     levels of any toxic air pollutant  for which  an emission factor can
    rmined (e.g., PNA's and aldehydes) to be estimated.
         ient
          for this study was provided by the U. S. Department of Energy,
     Northwest and Alaska Regional  Biomass  Energy Program, administered
    B°nneville Power Administration.
  jj s- Environmental  Protection Agency,  Compilation of  Air Pollution
  ^jsion Factors.  Section 1.10 - Wood  Stoves,  Research Triangle Park,
    1  December 1977.

   '    Shelton and L.  W.  Gay, Evaluation  of  Low-Emission Wood Stoves.
        rnia  Air  Resources Board  Report  (Contract A3-122-32),  Sacra-
        CA,  1986, 109  pages.
  Q
  \f' A- Simons,  P.  D.   Christiansen,  J.  E.  Houck  and L.  C.  Pritchett,
          e  Samoline  Methods Comparability  Analysis  and  In Situ Evalua-
          New Technology Woodstoves  - Task G  (Draft), U.  S.  Department
     Energy,  Pacific   Northwest  and  Alaska  Regional  Biomass  Energy
          Report  (Contract  DE-AC79-85BP18508) ,  Portland,  OR,  1987,  58
        plus appendices.
                                  667
-------
  mi
                          •  11  it
    Figure  1.    Mean Hourly CO
         Eugene,  Oregon
         All Winter Days
(n=6268,  ordinate is in /*g/m3 CO)
   Figure 2.   Mean Hourly c°
         Eugene, Oregon
  Winter  Sundays  and  Solid*? oj
(n=986,  ordinate is in
       Table  I.   Temperature,  Traffic and RWC Impacts on Winter
                       CO




City
Boise
Eugene
Libby
Missoula
Portland
Yakima




n
1,668
6,268
866
4,253
15,274
1,393
Mean Total Hourly CO (Mg/m3)


All
Days
2.23
1.53
2.74
2.90
0.92
2.14


Coldb
Days
2.85
2.27
N.D.
3.71
1.07
2.36


Week-
days
2.38
1,65
2.94
3.09
0.98
2.34
Week-
ends &
Holi-
days
1.85
1.22
2.24
2.43
0.78
1.64
Mean Upper-Limit 3}
Hourly CO (^^-


All
Days
0.88
0.54
0.88
0.55
0.25
0.28


Cold.
Days
— —
1.51
0.96
0.78
0.73
0.34
0.36
ds*
it.
^° 1$
J^
l'l\
, «5
0-°
. t1}

°' 
-------
Figure  3.   Mean Hourly CO During Low Temperature Period (HDD = 63),
     Christmas 1985 - Boise, Idaho (ordinate is in /ig/m3 CO)
Figure 4.  Mean Hourly CO During Mild  Temperature  Period (HDD - 9),
     Monday 2/24/86 - Boise, Idaho  (ordinate  is in  /*g/m3 CO)
                                  669
-------
UTILIZATION OF CARBON-BASED ADSORBENTS FOR MONITORING
ADSORBATES IN VARIOUS SAMPLING MODES OF OPERATION
William R. Betz and Matthew C. Firth
Supelco, Inc.
Supelco Park, Beliefonte, PA  16823-0048
     Organic contaminants in ambient air or other atmospheres can be
difficult to identify and quantify.  Lack of information about the
adsorbates, such as molecular weight, vapor pressure, surface activity,
etc. typically makes it necessary to use several sampling devices (i.e.,
adsorbent tubes), each adsorbing a different fraction or class of the
organic compounds present.  In contrast, carbon-based adsorbents are
specifically tailored to adsorb and subsequently desorb a wide range of
organic contaminants.  Use of these adsorbents, which include carbon
molecular sieves and carbon blacks with surfaces graphitized to
differing degrees, minimizes the need for more than one sampling device
and allows analysts to develop effective adsorption/desorption schemes.

     Carbon molecular sieves are the carbon skeletal frameworks
remaining after pyrolysis of synthetic polymers or petroleum pitches.
The physical characteristics of a sieve (surface area, micropore
percentage, etc.) are a function of the precursor and the employed
manufacturing parameters.  Hence, the physical characteristics of a
sieve can be tailored to permit the sieve to function in either general
or specific sampling applications.  These materials are usually used to
adsorb volatile organic adsorbates.
                                    670
-------
te   -*«fnj.uis!niu carbon blacks are pyrolyzed at extremely high
  Peratures  that allow the carbon surface to be rearranged into a
M^lte lattice structure.  As with carbon molecular sieves, the
    cal characteristics of graphitized carbon blacks can be tailored to
    general  and specific applications.  These materials are usually
    to adsorb semivolatile and nonvolatile adsorbates.

    Characterization and tailoring of carbon molecular sieves and
          carbon blacks have led to construction of adsorbent tubes
tub b0th Seneral and specific applications.  These include multi-bed
•  es for monitoring wide ranges of organic contaminants, and single bed
     for monitoring specific adsorbates such as acetone, methyl ethyl
     > 2,3-dibromopropanol, or ethylene oxide.  Additional applications
      developed to take advantage of the characteristics of these
    Volatile,  semivolatile,  and nonvolatile organic contaminants in
   6llt  air,  stack gas effluents, and indoor air atmospheres can be
   Jcult  to  identify and quantify.   Since these sample matrices
   -ally  contain a wide range of adsorbates—in many molecular sizes
flu 8hapes—choosing appropriate sample concentrating media becomes
Con~JCult. This choice of sample preparation media is further
Ut-M  ated if  the adsorbates possess one or more functional groups.
*li   tion of  Class I, nonspecific adsorbents, in an adsorbent tube,
poSgillates the  concern over which functional group(s) an adsorbate
M8        Use of an adsorbent tube containing several Class I
and°rbents» having different surface areas, allows for the adsorption
4(js 8ubsequent  thermal or solvent desorption of a wide range of
  °rbates possessing different molecular sizes and functional groups.

    BV  definition,  a Class I adsorbent is without ions or active
        •. -   > .   |  -  j,    -,       Mi
 ou   '   Only  a Class  I adsorbent interacts nonspecifically with the
U0,  groups  of adsorbates,  ranging from the n-alkanes (group A
H0ucules) to  the  alcohols,  organic acids,  and organic bases (group D
at *Cules).l  Graphitized carbon blacks,  and some carbon molecular
a  are  classified as Class I, nonspecific adsorbents,1  The
^tbrptive Pr°Perties  of three graphitized carbon blacks and several
^t]?n molecular sieves were characterized by following procedures
evajllle
-------
were introduced into the tube via typical gas chromatographic injectio
techniques.  The specific retention volume—or breakthrough volume--
the calculated volume of gas needed to cause the "challenge slug" 0±baclc
injected adsorbate to migrate from the front end of the tube to the
of the tube (Table I).  The adsorbates, including water and other P° ^
molecules, were chosen to provide insight into the classification or
adsorbent surface.

Results
     Characterization of the carbon molecular sieves focused on
interactions with volatile organic contaminants, since the working
of a typical carbon sieve is the C2-C5 hydrocarbon range.  By
definition, carbon molecular sieves, or porous carbon blacks, are t
carbon skeletal frameworks remaining after thermal degradation of
synthetic polymer or petroleum pitch precursors.3»^  The physical    ^
characteristics of a sieve, such as surface area, pore distribution,^^
pore diameter, are functions of the starting materials and manufact
parameters employed.

     Utilization of a physical descriptor, such as the carbon:hydrog^
ratio, as the input value, provides insight into the percentage of
of surface functional groups (these can be desirable or undesirable
constituents) and sieve adsorption strength (Figure 1).  A sharp      p
increase in the C:H ratio from approximately 22 to 82 indicates a sn^
decrease in surface polarity, and that micropore closure has occurre
Pyrolysis of a carbon molecular sieve, at appropriate temperatures,
provides a class I, nonspecific sieve with excellent adsorption and
hydrophobic properties.

     The adsorptive and hydrophobic properties of a pyrolyzed carbon
molecular sieve are compared to the properties of activated charcoa
(Table II).  The specific retention volumes of dichloromethane for
Carboxen™-569 sieve and activated charcoal are similar.  The specif1
retention volumes of water, however, differ significantly, because
carbon oxides are present on the activated charcoal surface.  These
oxides are formed during the manufacturing of the charcoal, which    ^
entails intrusion of compressed air or steam under high  temperature
pressures.  The activated charcoal is, therefore, hydrophilic compa
to the Carboxen-569 sieve.

     High temperature pyrolysis of the carbon sieves produces an
ordered, aromatic carbon surface.  Other  starting carbons, such as
carbon blacks, also can readily undergo aromatlzation at high        g
temperatures.  As with the carbon sieves, the physical characterise
of these graphitized carbon blacks can be altered by varying the
manufacturing procedures.  For example, graphitized carbon blacks,
having surface areas  of 3, 10, and 100m2/gram, can be used to
construct  a multi-bed adsorbent tube  that will adsorb and thermally^,
desorb compounds  ranging from benzene  to  the high molecular weight
(Table III).  These graphized carbon  blacks are  also Class I         ^
adsorbents.  Water  adsorption becomes  a  function of molecular  size'tlle
competition for the adsorbent surface  is  won by  adsorbates such  as
aromatic  hydrocarbons.
                                    672
-------
 Iji  Characterization evaluations for these three graphitized carbon
 tet    indicate trends associated with Class I adsorbents.  The specific
 the J^*011 volumes of benzene, methylbenzene, and ethylbenzene illustrate
 (ja.   ect relationship between surface area and adsorbate retention
   ^e TV).  An increase in adsorbate surface area, with the addition of
        group, produces a linear increase in the specific retention
     3 for the three graphitized carbon blacks.  If the Carboxen-569
 l>, -"" molecular sieve is included with the three graphitized carbon
 - ^ks in a muiti-bed adsorbent tube (Carbopack™ F, Carbopack C,
         B, Carboxen-569, in order from the tube inlet end), the
       range of the resulting four-bed tube is expanded to include the
 ^5 straight chain hydrocarbons, as well as the aromatic
 aj8t0cart>ons.  In other words, the working range of this tube includes
  Orbates ranging from vinyl chloride to the PCBs.

    The hydrophobic properties of these adsorbents enables an analyst
      the four-bed tube to adsorb volatile, semivolatile, and
      tile fractions sparged (steam distilled) from aqueous sample
 fja   es«  The chromatogram in Figure 2 shows a sparged base-neutrals
 In Ction (110°C sparge vessel temperature) adsorbed on the four-bed tube
 ^ta 8ize-exclusion mode of operation.  Because approximately 2ml of
 teJr Pass through the tube during the sparge cycle in this high
 ls PSrature purge and trap analysis, the use of hydrophobic adsorbents
 f]_J*SSential.  The trapped analytes are thermally desorbed via typical
  ^reversal techniques.  Recovery ranges from 83 to 100%.  The tube
    successfully adsorbs and thermally desorbs these analytes in
   ent air sampling operations.

    Another Class I adsorbent,  Carbotrap™, has been used to adsorb
    'rne C4-C8 aliphatic amines (Figure 3).  Recovery of these
   ytes by chemical desorption ranged from 92 to 103%.   Carbotrap
   ~''ent also has been used in a multi-bed adsorbent tube containing
 a^ce adsorbents (in order from the inlet end): Carbotrap C, Carbotrap,
 to .£arbosieve™ S-III,  a carbon sieve with adsorptive properties similar
 dee  °8e of Carboxen-569.  This tube adsorbs and subsequently thermally
 In ^bs the airborne contaminants cited in Methods TO-1, TO-2,  and TO-3
H0 S EPA document #600/4-84-041, the compendium of methods for
 bg toring toxic organics in ambient air.  Figure 4 shows this tube can
^(j^cessfully combined with a gas chromatographic analysis of the TO-2
the    3 adsorbates,  by using a thermal desorber that directly transfers
8'  adsorbates to the chromatographic column,  without cryofocusing.   An
• * 1/8" stainless steel column packed with 1% SP -1000 on Carbopack B
        for this analysis,  with the temperature programmed from 35 to
 Oj.  c*rbon-based  adsorbents  have  been used  in various sampling modes,
 n_,Monitoring  organic  adsorbates.   The adsorbents studied,  which
       carbon  molecular  sieves  and  graphitized carbon blacks,  have been
      J to  function in defined  working ranges.  The Class I,
       fie  characteristics of these adsorbents allows them  to  perform
     ively  within these  ranges.
                                 673
-------
References
1.   A. V. Kiselev and Y. I. Yashin,  Gas Absorption Chromatography.)
     Plenum Press, New York, NY.  1969.

2.   U.S. Environmental Protection Agency,  Characterization of
     Resins for use in Airborne Environmental Sampling.   (EPA Doc
     #600/7-78-054 / NTIS Document #PB284347), National  Technical
     Information Service, Springfield, VA.   1978.

3.   R. Kaiser, J. Chromatographia 3; 38-40 (1970).

4.   J. Vi. Neely and E.G. Isacroff, Carbonaceous Adsorbents for Jtbje
     Treatment £f_ Ground and Surface Waters Macel  Dekker,  Inc., New
     York, NY. ~1982.

5.   E. R. Cropper and S. Kaminsky, J. Anal. Chem. 35;  735-743 (1963)-
                                   674
-------
                               Table I

       Calculations for Determining Specific Retention Volume
yt         milliliters of gas needed to cause adsorbate to migrate
 8  •      	__	
                     weight of adsorbent bed in grams
                    (tr-ta)
         (j) (Fc) [ _ ]
                        Wa
            3         (Pi/Po)2 -1
    J  -  _    t - — ]
            2         (Pi/Po)3 -1

                 Tc           Pw
   Fc  - (Fa) [ _ ] [ 1 - - ]
                 Ta           Pa
    J  "  Pressure correction factor
   ^c  B  Corrected flow rate
   fcr  =  Peak maximum retention time (apex) at equilibrium point
   ^a  =  Dead volume retention time
   ™a  •  Adsorbent weight
   ^i  *  Inlet pressure
   ^°  "  Outlet pressure
   ^a  •  Flow rate at ambient temperature
   ^c  a  Column temperature
   ™a  =  Ambient temperature
   ^  "  Vapor pressure of water (at "flow meter" temperature)
   **a  "=  Ambient pressure
     It is reported that j approaches unity in the pressure and
     temperature ranges evaluated. ^  Consequently, it has been
     assigned a value of 1 for this study.
                                 675
-------
                               Table  II

     Specific Retention Volumes for Dichloromethane and Water on
            Carboxen-569  and Activated Charcoal Adsorbents

                     Specific Retention Volume (ml/gram)
Adsorbate            Carboxen-569     Activated Charcoal

Dichloromethane      1.73 x 105       1.57 x 105
Water                2.24 x 102       9.76 x 103
                              Table III

        Characteristics  of  Graphitized  Carbon Black Adsorbents

                     Surface Area     Functional Range
Adsorbent            (ia2/gram)        (Molecular Size)

Carbopack F          3                >C20
Carbotrap C          10               C12 - C20
Carbotrap            100              C6 - C12
                              Table IV

             Specific Retention Volumes for Adsorbates on
                 Graphitized Carbon Black Adsorbents

                     Specific Retention Volume (ml/gram)
Adsorbate            Carbotrap        Carbotrap C   Carbopack F

Benzene              2.59 x 103       1.99 x 102    1.06 x 102
Toluene              3.35 x 104       7.77 x 102    3.93 x 102
Ethylbenzene         7.49 x 10*       1.64 x 103    6.87 x 102
                                   676
-------
                                 Dichloromethane
                           4   20  80  82   84
                             Carbon: Hydrogen Ratio
       1    Specific retention volumes of dichloromethane and water as a
      on  of  the  carbon:hydrogen ratio on the adsorbent  surface.
        0   '4   '   8  '   12  '   16   '  20     24  '  28
                               Min.
a mu       Base-neutral  organic compounds sparged from water, trapped  on
  UJ-ti~bed adsorbent tube,  and thermally desorbed to a GC column.
                                 677
-------
          n-Butylamine  Triethylamine (int. std.)





\


L_



L
n-Amylaminf


Triel
n-Hexyla
n-Hepty
n
Uj
hylamine
                                               11 ng Each Amine
                                 n-Octylamine
                                   Blank ORBO-100 Tube
                  I

                 2
4    6    8    10
  Min.
Figure  3    C^-Cg aliphatic  amines from an air  sample (solvent
desorption).
                      Vinyl Chloride
                                     Dichloromethane
                                I     I     I     I    I     I     I
                                8        12        16        20
                                    Min.
Figure 4    TO-2  and TO-3 compounds (thermal desorption).
                                678
-------
   of a Sampler for Peroxyacetyl Nitrate Monitoring
  Fung
       a Acaso
       CA 93010
|^JJeroxyacetyl  Nitrate   (PAN)  is  an  atmospheric  pollutant
        a result of photochemical  reaction of hydrocarbons and
  ,t --  provide  simple  and   cost   effective,   large  scale
  Cof1115  of PAN in  remote areas,  a sampler  has  been designed
  iiit ect  pAN  as  acetate  after   hydrolysis in  an  alkaline
       Ambient   acetic   acid   which  interferes  with   the
     ent  is removed  with a selective  scrubber.   Laboratory
       of the acetate content  of  the medium will  provide the
           amount of PAN collected.

       sampler  is   being  evaluated   in   terms   of  scrubber
     cy  and specificity,  PAN  collection  efficiency,  as  well
       potential   interferences   such  as   from   ozone   and
ly  .l
                           679
-------
Introduction

                                                          b
     PAN is a  secondary pollutant formed photochemical ly
reaction between  hydrocarbons and NOX  in the atmosphere.
it  was  first  identified  and   studied  by   Stephens
coworkers1'2,    PAN has been  measured in  many parts  of the
in  both polluted  and  unpolluted atmospheres-*"10.  PAN
transported over long distances4 '5,7 f 10  an(j nas t,een  suggeS
a  reservoir  for  NOX in  the troposhere11.  The  role  of  ^
precipitation chemistry has also been investigated12'1  •
     The most widely used technique for PAN measurements
gas  chromatography  with  electron   capture   detection/
provides  good  sensitivity  and   specificity.   A  diff  ^
technique   incorporating   an   alkaline   scrubbed    &
chemiluminescence   has  also  been  attempted14 .  Insti"
calibrations  have  been   generally  performed   usin9
atmospheres generated by  dilution  of a source  of synthetic
whose absolute concentration has been  established using ^  .
spectroscopy15, a NOX analyzer16, or by  determination  of n * j UJf
after hydrolysis  of PAN  in  alkali17.  PAN  has  been P*£ga  & *
irradiation  of  either   ethyl  nitrite in  dry  oxygen  / *^°
chlorine-acetaldehyde-NO2 mixture  in air8'19.  The  last t
was  utilized  to produce  a  portable  PAN  generator  f°r
calibration  of PAN  analyzers^0.  PAN has been  also
high yield by nitration of peracetic acid,  followed by
into  n-heptane17  and  purification with  high  pressure  eg
chromatography, or extraction  into  n-tridecane  and used  o1
     Instrument calibration represents the most difficult
the PAN measurement procedure.  Because  PAN is  thermally
and  easily lost  to  surfaced,  calibration  atmospheres
freshly prepared  and used.  PAN's high  reactivity also  P
handling problems. Pure  PAN liquid is highly  explosive.  ™
of the PAN generator has overcome some of these diff iculti6 '
the life  of the permeation tubes  used  in the system  ^-
long-term  output  stability.  Other reported  difficulties
preparation  of  columns   of  consistent  performance2^  <
degradation10,  adsorptive  losses within  the  instrument2  /  g£j
of detector response due  to  moisture23'24,  and column  JjTjty »
Part of these problems are related to the thermal  instafci1  j a
PAN.  For example, the GC oven and  detector must  be maintai
close to ambient temperature  in order to prevent signifa
decomposition during  the  analysis.   Consequently, the
foil is subject to contamination by column bleed.
                                                          of * 1
     Recently,  a PAN instrument based on chemiluminescence ^ PjJ
(from thermal decomposition of  PAN) with luminol is market'  ^Oo
Scintrex    (Toronto,   Canada) .  The   unit  offers  ver^j f
sensitivity. However, there has been  very limited  reP°rte^e(J
experience  with  this  instrument.  Evaluation   is   nee7 \
establish  the  specificity, stability,  and both short &rf
term performance of this instrument.  Judging  by the  Per
of the LMA-3  instrument ( for N02)  from  which the PAN
                               680
-------
       this PAN instrument could suffer  from  similar  problems,
    as  temperature dependence  and  nonlinearity  of  instrument
   nser  as  well as some hardware reliability.

       Operational Evaluation  Network (OEN)  is  a field  study
      y the Electric Power Research Institute to  collect  daily
      concentration data of various species in the eastern  and
Hj estern  U.S.  over  a  two-year period  for the  evaluation  of
Si*. ^Position  models.  PAN  is  one  of  the  species   being
  ^o
  s    .,  term,  large  scale monitoring  of PAN  in remote  areas
  3?nt problems  that  are  not  normally  encountered  in  small
  ;_es-  In the past, many studies were conducted  by  researchers
   elves,   or  technicians   under  their  supervision   in   a
   |-Vely  short time frame. In PAN measurement, calibration and
      ice are  essential  to  produce  data of  consistent high
      The chance  of having equipment malfunctioning or  drifted
f!te f\L-Calibration is much smaller  in  a short study than one like
       Thus  it is critical to have  qualified station operators
       capable of performing these tasks on a  routine  basis.
  h  .  such  technicians  are  few, if not  at  all, available  in
  network due  to the location of  the stations.
         are  also  other  considerations. The  equipment  costs  are
     for   ECD-gas  chromatographs  with  automatic   injection
       integrators,   and  on-site  calibration  systems   (PAN
 sis  °rs)/  which have yet to be proven reliable on a  long term
   j  and  which can  be  operated  by  a  low  level  technician.
   ies  such  as carrier  gas,  fittings,  columns, and  other  GC
    must  be  stocked,  or mailed  to  the  stations  regularily.
    ¥'  to conduct  low level ("0.1-4 ppbv) PAN measurement with
    Accuracy  on  a  continuous  basis  with  twenty eight   PAN
   ""•a  operated  and  maintained  by  twenty eight   different
     who are  most likely  nontechnical,  is almost impossible,
     the other constraints  are not  present.
      Design

      overcome these obvious difficulties,  a sampler has been
       to  collect  PAN  as  acetate  after  hydrolysis  on  an
       medium.   It  has  been  established   that  PAN  readily
         into acetate and nitrite  in the presence of  alkali2.
      	on  was used  to establish  the concentration  of PAN
  th   ally Produced17,  the output of  the  PAN generator20,  or
I'^yT  ibasis  of  an  indirect  measurement  of PAN  with  a NOX
N *   • witn this approach, much of  the burden of producing
         is  shifted from the field  to the  laboratory,  where
        collected  will  be   analyzed   to  determine  the
            amount  of  PAN.  As  such, the measurement  is much
         and  less  costly.   Other  benefits  include   better
   y  control,  reduced   field operator involvement  and lower
       maintenance.
                            681
-------
                                                        ..  and
     There  are  potential  interferences.   Acetic
particulate acetate in ambient air must be removed p]
collection of PAN.  Otherwise,  positive  bias  to the     ^
will  result.  Particulate  acetate  in  the  sample  air    t
removed with  a  Teflon  filter.    To  remove  acetic  acid, ^
essential that the process, while being ^.^^iYJnts  £°r
acid, must  leave  PAN  intact.  The Henry's Law constants',
and  acetic acid  are  in  the  order  of  3  and  5000  m
respectively  at  25°C12.  At  pH  7,  the  apparent  Henry
ico^c      jf               -0-77.nnn.   An  advantage  is  *••» .
respec                 .
constant for  the  acid is  '877,000.   An advantage ,
?£is huge solubility difference between PAN and  acet ic    » rtf
the  scrubber.  Using  nearly  ice  cold "^^  af  ™selightiy
solution,  the solubility of PAN  is  only  increased slig    all
V-  1  M/atm),  but  its  decay  in  solution   is  signi
retarded" ,   thereby  reduces  the  amount  of  PAN  l°s* iU
system. If  sampling  is  conducted at  l  l^er/min   it  w^l
take  the  first  few liters of  air  containing  PAN at  *   $
satisfy  the  solubility  requirement  in  the  scrubber ;
sampling  will be  conducted  for  24  hours,  this  loss
considered as negligible.

     The large solubility constant  for acetic acid means '
can  be  removed  readily with  the  water  scrubber •    Th*A
temperature used should further enhances  the  ef f ^Jnc
impinger  has  been  shown  to  collect  acetic  acid
efficiency previously25.

      Other  potential  interferences have  been  also

     sssr7ssss~of Kea^r^v^^          js

                             of Tcet^Td^de^y
produce acetate have also been examined ^^SLn^O   *
to proceed  readily under the experimental  c°"dlt^ldehyd
sampler,  it may be  expected  that  much of the ac etaldeW^
peroxides could end up in the  scrubber and little will r
alkaline  medium  which  collects the PAN   Nevertheless
atmospheres containing these various  in^6"^"
the  PAN  sampler  will be  performed to delineate  the
 interactions and any bias that may  result.
      in summary,  a sampler  has been designed to
 an alkaline medium.  Interferences f rom Acetate
 are removed with  filtration  and a cold  water  Rubber
 potential interferences such as from acetal^hy^n
 ozone and peroxide have been considered and are being
 Acknowledgement
                                                        E
      This research was conducted under funding  from the
 Power  Research  Institute  as  part  of   the   funding
 Operational Evaluation Network.
                               682
-------
'
   E. R. Stephens,  P.L.  Hanst,  R.C.  Doerr and W.  E.  Scott,
   Reaction  of  nitrogen  dioxide and  organic compounds in air
   Ina. Enaaancf.  Chem. 48:  1498 (1956).
   ETR. Stephens,  The formation,  reactions and properties of
   Peroxyacetylnitrates  (PANS)  in photochemical air pollution
       Enviro.  Sci.  21: 119 (1969).
   S-A.  Penkett,  F.J.  Sandal Is,  and J.E.  Lovelock, Observations
   °f peroxyacetyl nitrate (PAN)  in air in southern England
          Environ. 9:  139 (1975).
   W-A.  Lonneman,  J.J.  Bufalini, and R.L. Seila, PAN and
   °xidant measurement in ambient atmosphere Environ. Sci.
            10:  374 (1976)

   H.  Nieboer and J.  van Ham Peroxyacetyl nitrate (PAN) in
   Delation to ozone and some meteorological parameters at
   Delft in the Netherlands, afmns. Environ. 1Q: 115 (1976).
   H.B.  Singh, L.J. Salas, H. Shigeishi, A.J. Smith, E.
   Scribner, and L.A. Cavanagh Atmospheric distributions,
   sources and sinks of selected halocarbons, hydrocarbons,
   SPfi and N->0 EPA-600/3-79-107 U.S.E.P.A., 1979.

   T- Nielsen, U. Sammuelsson, P. Grennfelt and E.L. Thomsen
   peroxyacetylnitrate in long-range transported polluted air.
           Lond. 293: 553  (1981)
8
   H-B. Singh, L.J. Salas, Methodology for the analysis of
   Peroxyacetyl nitrate  (PAN) in the unpolluted atmosphere.
   ^fcaos. Environ. 17: 1507  (1983).

   C-W. spicer, M.W. Holdren, and  G. W. Keigley The ubiquity
   °f peroxyacetyl nitrate in the  continental boundary layer.
   &£ffios. Environ. 17: 1055  (1983).

 '  K-A. Brice, S.A. Penkett, D.H.F. Atkins and F.J. Sandalls,
   D-J. Bamber, A.F. Tuck and G. Vaughan Atmospheric measure-
   fcents of peroxyacetyl nitrate  (PAN) in rural south-east
   Sftgland: seasonal variations, winter photochemistry, and
   long-range transport. i^mns. Environ. 18.: 2691  (1984).
k
 '  H-B. singh and  P.L. Hanst Peroxyacetyl nitrate  (PAN) in  the
   ^polluted atmosphere: an important reservoir for nitrogen
j  °*ides. Geophvs. Res. Lett.  8;  941  (1981).

 ^  V.K. Lee, G.I.  Senum  and  J.S.  Gaffney Peroxyacetyl  nitrate
    (pAN) stability, solubility, and reactivity  - implications
   for tropospheric nitrogen cycles and precipitation  chemistry
   Fifth international Conference of the Commission on Atmos-
   *>heric Chemistry and  Global  Pollution, Symposium on Tropo-
   sPheric Chemistry. Oxford, England, Aug.  28-Sep. 2,  1983.
                             683
-------
13.  M.W. Holdren and C.W. Spicer, and J. M. Hales peroxyace Y
     nitrate solubility and decomposition rate in acidic wa
     Atmos. Environ. 18; 1171 (1984).

14.  D. Grosjean and J. Harrison, Peroxyacetyl nitrate:
     comparison of alkaline hydrolysis and chemiluminescenc
     methods. Environ. Sci. Technol. 19: 749 (1985).

15.  E.R. Stephens, Absorptivities for infrared deterrainati0
     peroxacetyl nitrate. Anal. Chem. 36; 928  (1964).

16.  L.F. Joos, W.F. Landolt and H. Leuenberger Calibration
     peroxacetyl nitrate measurements with an NOX analyze*"'
     Environ. Sci.Technol. 20: 1269 (1986).

17.  T. Nielsen, A.M. Hansen and E.L. Thomsen,  A conv
     method for preparation of pure standards of
     nitrate for atmospheric analysis Atmos. Environ.
     (1982) .

18.  E.R. Stephens, F.R. Burleson and E.A. Cardiff, JA£C3' ^**
     (1965).

19.  B.W. Gay, Jr., R.c. Noonan, J.J. Bufalini, and P.L. ^a*1 ^
     Photochemical synthesis of peroxyacetyl nitrate in 9aS.g. *
     via chlorine-aldehyde reaction Environ. Sci. Technol^ —"
     (1976).

20.  D. Grosjean, K. Fung, J. Collins, J. Harrison and E. ^f
     Breitung, Portable generator for on-site calibration
     peroxyacetyl nitrate analyzers Anal. Chem. 56; 569
21.  J.S. Gaffney, R. Fajer and G.I. Senum, An improved     :
     for high purity gaseous peroxyacetyl nitrate production-
     of heavy lipid solvents Atmos. Environ. 18; 215  (1984)'

22.  S.A. Penkett, F.J. Sandalls and B.M.R. Jones, PAN
     measurements in England - analytical methods and
     VDI-Berichte Nr. 270: 47  (1977).

23.  M.W. Holden and R.A. Rasmussen, Environ. Sci.
     185 (1976).

24.  I. Watanabe and E. Stephens, Reexamination of moisture  ,
     anomaly in analysis of peroxyacetyl nitrate. Environ^—*^
     Technol. 12: 222  (1978).

25.  K. Fung , Measurement of atmospheric organic acids:
     erations regarding sampling artifacts and potential
     ferences. Proceedings of the 1987 EPA/APCA Symposium o
     Measurement of Toxic and Related Air Pollutants, Ralei9
     NC., pp 208-211.
                               684
-------
       nCE EVALUATION OF THE HARVARD/EPA
      -ENUDER SYSTEM UNDER SIMULATED ATMOSPHERES
      Brauer. Petros Koutrakis, James L. Slater,
     J. Keeler, Jack M. Wolfson, and John D. Spengler

JSS^4 University, School of Public Health
 °str>Utltington Avenue
   °n. Massachusetts 02115, USA
          of  laboratory studies  to  evaluate  the  Harvard-E.P.A.  Annular
      System  (HEADS)  indicated excellent performance  under  a wide range
      ated conditions.  The HEADS consists of a glass  inlet/impactor, two
     coated  annular  denuders  to  collect gaseous  S02, HN03  and  HN02,   a
      acid-coated annular  denuder  to trap  NH3,   and a  filter  pack to
           particles  and artifact gases.  The inlet  is unique  in that  a
         impaction  plate  allows  the   inlet  walls   to   be  extracted.
          experiments   with  S02,  HNOj  and  NH3  vapors   at  various
      ations and relative humidities  (10-90X)  indicated that the annular
     8  achieve  collection  efficiencies  near  100X.  Similar  tests  to
     gate the collection of HN03 and  NH3 on  the walls of the glass inlet
     ^ small loss (<10X).  By extracting the first denuder and the inlet
         HN03 and NH3 can be recovered.
                                  685
-------
 INTRODUCTION

      The  Harvard/EPA  Annular  Denuder System  (HEADS),  designed  -   pl.
 sampling of atmospheric  aerosols  and gases,  is presently being used *" jo
 Acid  Aerosol   Effects   in   North  American   Children  Study.  pri°Li 1"
 utilization of  the sampling  system  in this extensive  field study, *" it
 conjunction  with   an  atmospheric  sampling   evaluation,   we  sougn tofy
 characterize  the   performance  of  the  HEADS  in  controlled  I*"30 -joU*
 settings.   Furthermore,  since  atmospheric  sampling is subject to  nU0,jj *
 artifact-producing  reactions,  the  evaluation  of  gas  collection   tS,
 controlled setting  allowed us  to  investigate  the sampling  of pure
 the  actual measured  endpoints.   To  evaluate  the  collection of  I
 species by the annular denuder portion of the system, we conducted  a
 of tests  at various  gas  concentrations  and relative  humidities.
 laboratory experiments enabled us  to sample particle-free air, we 1
 to quantify  the extent  of gas  collection on  the inlet  surfaces     ,„
 non-specific adsorption  processes.  The  results of  these  experiment*
 presented and discussed in this paper.

 SAMPLING AND ANALYSIS

      The HEADS  includes  a glass  impactor,  three annular  denuders &  •*
 filter pack.   As we have previously presented the design and  "  ~~"
     er pac.    s we  ave prevousy presented te design and descrm*
 this sampling system1,  and since the evaluation  of  the impactor and,il  b'
 pack  components have  been  reported  elsewhere1 «2'3,   this  paper  wl* tl>8
 limited to  a  description of  the  collection of  gaseous  species
 denuder and  inlet  portions of  the system.  Air  samples enter  the
 through the  glass  inlet/impactor  section.   Since  the  removable
 plate is connected to the first annular denuder  the glass  inlet
 extracted   to  recover  any  gases  which  adsorb  on  the  inlet
 Additionally, by removing the  impaction plate the first annular
 be  extracted  without any interferences  from coarse particles
                                                             '       * •
     The acid  gas denuder is coated with  10  ml of IX (w/v) NA2C03,  *•*  m\
 glycerol in a  1:1  methanol/water solution  to  collect S02,  HN03, *° o  «  in
 The NH3 denuder is  coated with 2%  (w/v)  citric acid, IX  (w/v)  glvc£ t\*\
 methanol.   After coating, the denuders  were immediately dried  wi*   $
 dry air and capped  to protect  them from acidic gases and NH3.   All »e"
 and  inlets  were extracted with 10  ml  of  ultra-high purity  ^&tl.lc
 analyzed by  ion chromatography  on a  Dionex  4000i.   Limits  of °"e .*
 (LOD)  were  based upon extract concentrations of 0.1  ^g/ml,  which **
 half  the   concentration  of  the   lowest  level  standard  used  *p
 construction  of standard  curves.
                                                                      »,
    We   constructed  an   experimental   gas   generation   systeffli  t fflt
 diagrammatically in Figure 1, in order to  evaluate  the collection °^eA
 gaseous  pollutant species, NH3   S02 and  HN03.  S02  and NH3  were B^
 from  commercially prepared gas' cylinders  while gaseous  HN6, was &e**
by  nebulizing dilute  HN03 solutions.  All  dilution air in the  system V
 through  a  series  of scrubbers:   Purafil  to  remove  nitrogen  °*^lt
ozone,  citric acid-coated glass  wool to  remove NH3,  GI activated  ctl*6c
 for  further removal of reactive   gases, soda lime to remove acidic s" &
and  a  0.5  pm  filter  for  particle  removal.  Sampling  flow  rate5
maintained at 10 Lmin'1 by mass -flow controlled pumping units and  &*•
rates were measured with rotameters.  All  tubing  in  the system was **
teflon  (PTFE).'S02  and HNOs collection on NAgCOs coated denuders   »
              .2  an     s coecon  on    gs  coate   denuers
investigated  using  the  Test Atmosphere Generation  System  (TAGS)  *  ^
laboratories   of  ERT,   Environmental   Consulting   and  EngineerioB' 
-------
   's AND DISCUSSION

        HN03  and   NHr   collection  efficiencies   were  determined  at
                an     ,  coecon   ecences   were    etermined  a
             10  and 90*  m-  Results  are  shown in  Table  I.   In  each
        '  two  or  three systems,  comprised  of  an  inlet  and two  coated
    1  l sampled  simultaneously from the same  manifold.   Based on  the  10
  r cn«     rate,  the detection limit per hour  of sampling corresponds  to
  >ct?,Centratlons  of  °'4'  °'7  and  °-3  PPb  for   s°2-  HNO,  and  NH,,
  JNt  | y-  Collection  efficiencies  were  calculated  by  dividing  the
  , 1W °* Sas  collected  on  the  first  denuder  by   the  sum  of  the  gas
  'Jond ~ °n the  first and  second  denuders.  If  no gas was detected  on the
  > lcien   uder   ln  the  samp!ing   train,  the  experimental  collection
  ** col] y was determined by using the LOD as  an estimate of the amount  of
       0    b?  the second denuder.  Laboratory  denuder blanks collected
             levels  of gases.  NH3 concentrations  were  calculated  from
                ,              .         2  exracs  were oxze   y ang
      i   a 3Z a(Iueous H202 solution.  The addition of H202 is necessary to
          oxidize  S03- ,   formed   from  S02  and NA^COj,  to  S062'  for  ion
        aphic  analysis.  In  atmospheric   sampling,  the   presence  of
        sPecies  in  the  air  sample is usually sufficient  to completely
        the °°llected S02 to  S042'.  The  results  of  the  efficiency tests
    av  above  are  indicative  of experimental  collection  efficiencies
 til?9 fo  "early equivalent to the predicted  collection  efficiencies  of
  *6CM  tne three gases.1-4  Additionally,  we  found no evidence  that the
    *-l°n efficiencies are affected by RH.

  ltvuB *mP°rtant issue  to address  in  the  evaluation  of any  gas  sampling
  Ur  1 is  the  non-specific collection  of  gases  on  surfaces  of  the
   Coii    particular the  sampling inlet.   It  has been demonstrated that
  ed t,>e^tion on inlets  commonly  occurs but is  decreased when inlets are
  Ssp  h  teflon (PTFE).5 Furthermore,  the collection of HN03  on inlets
  S **s  at high relative humidities.5  We were interested in testing the
  ecti  these effects on HN03 collection and  in  determining whether NH3
  • N   °n the  inlet would  °e  increased  if  teflon-coated  inlets  were
     j'J  f °2  l°ss on  the sampling  inlet was observed under  a wide  range
     on  les  and 8as  concentrations.  Experiments  to evaluate the effect
        inlet  collection,  shown in Table II, were performed  both  at
      *nd  at ERT.  Additionally,  HN03 and NH3 collection  on  teflon- coated
   and  ?d  inlets were compared.   To  evaluate  the inlet  losses, denuders
  8ame  *tnout  connected inlets  (uncoated or teflon-coated glass)  sampled
     e
  fcct»H     stream-   Inlets  were extracted to determine the  amount  of NH3
        n  the  inlet-   Results of these  inlet  tests,  shown in Table  II,
         cate  that "^ and ^3 losses  on  the inlet are  minor. In  all
      uf it:  was demonstrated,  that  following extraction of the inlets,
      HNOj  or NH3  could be  quantitatively recovered.
 the6 inlet and denuder components of the HEADS are' shown to be suitable
S ga  "easureraent of  atmospheric  gases.   Laboratory experiments  using
   otS resulted in mean  collection efficiencies  of 0.999,  0.992  and
       S°2*  HN0?-  and  NHj,.  respectively.  These  efficiencies  compare
      Predicte° values.  Inlet  losses  of HNOs and NH3  were  small  but
    em.                 .                              3
  6<>tBH   to  demonstrate  the  ability  to  quantitatively  extract  the
H) ^co  material  from  the inlet.  Mean loss  of HNO, on  inlets  was 5.8Z
Vb (u»T  d  6lass  inlets and  1.3X  for teflon-coated glass  inlets.  Mean
  Vert Coated and teflon-coated inlets) loss of NH3 was 2.9%.  No  S02 was
  thrn ln inlet extract samples.  It has  therefore  been shown that gases
     ?u8n the glass inlet  with only minimal adsorption  losses  into the
     Qenuder where they are absorbed completely by the denuder coatings.
                                  687
-------
ACKNOWLEDGMENTS

     This  work was  supported by  the National  Institute of  Environ^*($
Health  Sciences  (NIEHS) grant no.  1R01  ES0495-01,  Electric Power K«*b jjitf
Institute  contract no.  RP-1001,  and  the  Department of National Heal ^5
Welfare Canada,  Environmental Protection Branch.  MB is  supported by  jefl
grant no.  ES-07155-03.   Special thanks  are due to  Steve Heisler &n°tSi »*
Harrison of  ERT  for helping us to conduct some of the laboratory ter p^
part of the  Harvard-ERT Intercomparison Study,  funded by  the Electric
Research Institute.

REFERENCES
                                                                     1 pf 1
1. Koutrakis, P.,  Wolfson,  J.M., Slater,  J.L.,  Brauer, M.,  Spengleff  '(
Stevens, R.K., and Stone, C.L.,  "Evaluation of  an annular denuder/tli
pack system  to collect  acidic aerosols and  gases,'
Environmental Science and Technology.

2. Slater, J.L., Brauer, M.,  Koutrakis, P.,  and Keeler, J.,  "Determi"*
of aerosol strong  acidity using  an annular  denuder  system with a new j
designed filter, pack,"  submitted to the Proe. of the 1988 EPA/AP-GA
Symposium on Measurement of Toxic  and Related Air Pollutants.

3. Koutrakis, P.,  Wolfson,  J.M., Brauer,  M., Spengler, J.D., and Steve
R.K., "Design of a glass impactor  for an  annular denuder/fliter pacK
system," submitted to the Proc.  of the 1988  EPA/APCA Symposium on
Measurement  of Toxic and Related Air  Pollutants.
                                                                     J.P"
                                                     Submitted to
4. Brauer, M., Koutrakis, P., Slater, J.L. , Wolfson, J.M. ,
and Stevens, R.K.,  "Evaluation of the Harvard/EPA annular denuder
system,"submitted to the Proc. of the 81st. Annual Meeting  of

5. Appel, B.R., Povard, V., and Kothny, E.L.,  "Loss of nitric acid^
inlet devices  for atmospheric sampling," P_rp_c_j. of the J.987
on Measurement of Toxic Air Pollutants. pp. 158-167.

Table I.   Collection efficiencies of HNO^, NH3, and S0?.
Gas

HNOj












NH3
J










S02

Sampling
Duration
(hours)
3
3
3
3
1 *
1 *
3 *
3 *
24
24
24
22
24
3
3-
3 tf
3 f
13 #
4 #
4*#
4 #
4 //
24
24
24
2
2
Z R.H.

11 0.4
85 5.0
12 0.5
87 5.0
9 0.5
83 5.0
10 0.5
85 5.0
50 7.0
50 7.0
50 7.0
50 7.0
50 7.0
11 1.8
85 3.3
25 10.0
25 10.0
25 10.0
25 10.0
25 10.0
25 10.0
25 10.0
50 5.0
50 5.0
50 5.0
90 5,0
10 2.0
1st Denuder
(ppb)

74.0
59.7
10.9
18.0
168.8
144.0
7.4
13.1
12.0
21.6
25.6
30.7
3.4
183.3
164.3
184.3
190.4
55.4
27.2
23.4
29.0
60.2
18.0
16.9
2.3
206,0
182.0
2nd Denuder
(ppb)

-------
^xJil- Inlet loss of HN03 and NH3.
nlet type
^
^*»
N- COATED
jggT
"^ I
jll:
§8 :
^B>
/^N. COATED
tdQV)
% R.H.
90 .5.0
90 .5.0
90 .5.0
90 .5.0
85 -5.0
83 .5.0
10 .0.3
10 .0.3
85 *5.0
85 .5.0
85 -5.0
85 .5.0
Denude r
(ppb)
38.9
42.5
47.5
53.8
52.3
55.7
47.2
51.4
12.9
13.3
140.9
143.6
147.6
7.2
7.4
7.7
168.9
168.6
169.0
7.6
8.4
16.5
15.9
26.4
26.4
29.0
29.1
Inlet
(ppb)
4.3
3.6
4.1
0.8
3.9
0.6
1.4
1.8
4.7
7.6
5.4
0.5
0.4
0.4
5.2
2.5
1.5
#
0.6
1.0
0.4
1.0
0.8
Inlet +
Denuder
(ppb)
43.2
42.5
51.1
53.8
56.4
56.5
51.1
52.0
14.3
15.1
145.6
150.2
153.0
7.7
7.8
8.1
174.1
171.1
170.5

17.1
15.9
27.4
26.8
30.0
29.1
% on
inlet
10.0
7.0
7.3
1.4
7.6
1.2
9.8
11.9
3.2
5.1
3.5
6.5
5.1
4.9
3.0
1.5
0.9

3.5
3.6
1.5
3.3
2.7
U^^je collected at ERT with TAGS
^<
-------
         Fig tare 1.
         Simulated Atmosphere Denuder Test System
Air Scrubbers
Purafil

Charcoal

Soda Lime

Citric Acid

Particle Filter
                     Silica Gel
          Gas Generation
         NH3  S02
                                  Humidification
ter
7


-V
H2
^S

«-*•'
0
^=
— _ ~-
1- ...... 	




J
1
Humidified Air

H20
Trap


Heating Tape

                                     Dry Air
                                                        Rotameter
                                          Rotameter
               f Capillarg
                        Air
                        Scrubbers
HN03
Nebulizer
•i^MP
F
V
/

                                         HN03
                                         Reservoir
nr\p

Pump
lOL/min
R.H. /Temp
erature I I

Denuders

	 ••••ItaBBa 	
                                 690
-------
    ON OF ANNULAR DENUDER SYSTEMS
             ACIDITY MEASUREMENTS
           THE NETHERLANDS
    Waldman and Paul J. Lioy
  j n6nt: of Environmental & Community Medicine
   * ollert Wood Johnson Medical School
    es Une, Piscataway, NJ   08857  USA

    Vatl der Meulen and Hans Reijnders
      Institute of Public Health & E
    °x 1, BA 3720 Bilthoven, THE NETHERLANDS
"cj0   " "et neuien ana nans Keijnaers
°- B,  *nstitute of Public Health & Environmental Hygiene (RIVM)
 4C ayfis
 •  E   ects Research Laboratory
 earn^ir°nmental Protection Agency  (MD-55)
    ch Triangle Park, NC   27711  USA
  et?  initiate   a  year -long  study   of  atmospheric  acidity  in   the
  de anc*s .   a  2 -week study  was  conducted  using two  identical annular
     5ystems.  The systems each  included a  Teflon-coated  impactor  inlet,
           denuders  followed by a two-stage filter pack.   The  first  two
      Were coated with sodium carbonate to  remove  acid  gases  (S02,  HN(>2,
      ^  tne  third denuder is coated with citric acid  to remove  gaseous
       filter  pack  contained Teflon membrane filter followed by  a  nylon
      . filter  to collect particles for  anion,  ammonium  and  acidity
Nt  ° complete  systems were  set up  side-by-side atop  a field  trailer
'87  e suburban  village of Bilthoven  and operated for 2 weeks in August
^du  Tne   sai»pling   schedule  was  designed  to  determine  monitoring
I'*, i °ibility  and to  ascertain the  effect of delayed sample changing,
6fote etting  the  exposed samples remain  in the field  for a limited  time
ik^sti ^ck  UP-    This  study   design  consideration   was  addressed   for
Ve  cal reasons  because the systems will  be  installed in remote areas,
   Set-up and pick-up might take place  during weekly  visits,

 !titt°0d to excellent  reproducibility was found in replicate operation of
    al  systems.   No  differences  were  noted  for  samples  which   were
  sof immediately compared to  samples  that remained  in  the field after
  *.  for  24 to 48  hours.   Apparently,  the  denuders served to  isolate
  c<«eTs'   ant*  diffusion  through  the  impactor to the denuders  is
          slow to prevent post-run contamination.
                                  691
-------
INTRODUCTION

     Investigations in  recent  years  have  raised concerns about the
of acidic  sulfate aerosols  and other  atmospheric  acids on  expose   ^
populations  (1).   To support  health  effects  research,  measurement   $
are  being  further  developed  and  deployed  to document  exposures
relevant atmospheric  species.   Techniques have  evolved from  sin&A
ponent analyses  toward  the complete  speciation of  the  gaseous and
components which contribute to or attenuate atmospheric acidity.
     The diffusion  denuder  are proving to be  an effective tool app
air  sampling  for   this  purpose.   With  an annular  design,  denude   Ol
quantitatively collect reactive gases, while allowing fine -fraction a  jpt.
to  pass  (2).   Gaseous  HN03 ,  HN02 ,   S02 ,  and NH3  can  be measured  * ,$
levels in  the same air  samples  where aerosols  are  collected.   Rein ^ &
these gases  upstream  of the aerosol  sample  also reduces artifacts   ^(
exogenic  sulfate  formation  or aerosol  neutralization  on the  filte ^$
strate.    This method  allows more   meaningful  accounting  of atrnoSPjjl)
constituents which  can move between gas and aerosol phases,  such &s
and N(V) compounds,
     A year- long survey of atmospheric acidity  in The Netherlands
in  August 1987  by conducting  side-by-side field  testing of  two
denuder  systems  (ADS).    The  two-week  study  was  designed  to
sampler reproducibility  and to evaluate various scheduling consider*
These  results  also preview  the  chemical  characterization  for   in°
acidity/alkalinity  constituents  in  the  region,  which contains  afl° e
most intense ammonia sources in Europe  (3) .

METHODS

                                                                    iX
     Sampler.   Two identical  annular denuder  systems  (ADS)  were °" ' fl
simultaneously  atop an  air monitoring  trailer  (inlet  -5 m  above  6 gj
level).   The  site  was  adjacent to an open  field on the grounds of c  *
in Bilthoven, a suburban district NE  of Utrecht and ~50 km SE of
Little  industrial  activity and only  modest vehicular traffic
the district.
                                                                      t<
     The  ADS  units each  consisted of  a  jet impactor  preseparato* • ^
denuder  tubes  and a  two-stage  filter  pack   (manufactured  by un  ^ "[
Research  Glassware, Carrboro,  NC) .   The  impactor  and denuders are .f0p '
Teflon-coated glass (4).   The  impactor passes  particles <2,5 urn;  a  g$
silicone  oil was used on the impactor plate to  prevent particle re-e ^j7
ment.   The denuders were  tubes with 34 -mm ID, 1-mm  annulus,  and &
long annular region followed by a 25 -mm cylindrical section.            ,
                                                                   , -nef kl
     Reactive gases deposit to the walls  in the  annular region   a
on  the  chemical coating applied to  tube.   Fine particles pass
tubes  with negligible  deposition (5).   The  first and  second  tube
coated  with  Na2C03 ,  and  the  third tube was coated with  citric  aci
25 -mm  recess  was situated  on the  downstream end of each tube in °
reduce  turbulence  (hence particle deposition)  at  the  union between
The  carbonate coating  leads  to collection  of acidic  gases,  na
HNOj , and  HN02 ;  the  citric  acid coating serves to remove gaseous
purpose of the  second carbonate -coated denuder  is  to  provide a c°
for  the small  (but  non zero)  amount of particles  and non-reactiv
(mainly N02?  which are  collected  with equal  efficiency  on both c^ ^
and second tubes.  Hence, the gaseous concentrations were calculate

                                   692
-------
 the „
 «a,jeuirt:erence of analytes values  for the two tubes.  The  filter pack was
 %pD    Teflon and held two 47-mm  diameter filters  on stainless steel mesh
 by    s:   a Teflon membrane (2-um pore size without  support pad) followed
   .Nylon filter with a spacer between the  two  stages.   The Teflon filter
       all   particulate   matter.     However,  under   certain  conditions,
       nitrate aerosol can  dissociate (to NH3  and  HN03)  off  the Teflon
 tiltr"'g    "^he  Nylon backup  filter  was  used  to capture  any  volatilized
 %   6-   ^e  sum °f analyte on the two  filters give an unbiased measure of
   aerosol nitrate.
tilf*  9mpl
    e
        e  flows  were maintained at "1 m h   using  a  diaphragm pump and a
     ent^ai-flow  controller.   Total  sample  flows were  determined using in-
       gas meters, calibrated with  a  primary standard flowmeter.
                     n.  Extraction and Analysis.   The daily  operation of
 Pith''"15  samplers  required extraction of exposed denuder tubes,  followed by
 ^5  . ation  of fresh ones.   The tubes were  prepared in batches.   Coating
 HH, G°ne with  20 mL of  solution--either  Na2C03  (1%)  or  Citric Acid  (1%)
 V*s ad        ^1%^  ln H20:Metnano1  (1;1)'   To apply  the  coating,  an aliquot
 8tve    d to a tube,  and both  ends  were   capped.   The tube was shaken to
 coat.even coverage, then  the solution was decanted.   A single  aliquot of
 *B»   S  solution  was used to coat all tubes  in a  batch, and an unused tube
    ^es_erved  for blank  correction.   The  freshly coated tubes  were  dried
           of ultra pure air at 2-3  mL min"1 and stored with  end caps in
V  ^e ADS  components  were assembled and disassembled in  the  laboratory.
\( °mplete  unit was carried to the  field,  placed in the field  housing,
HL&Tld returned  to  the  lab.   Following sample  collection,  the  denuder
^t0 .Vere extracted using two sequentially added and decanted  aliquots  of
'M g 2ed (DI) H20  (20 mL total).  The filters were removed from the  packs
Hth tored  in polystyrene petri dishes.   The Teflon  filters were stored
            acid  impregnated  paper   filters  to  provide  a  zero-ammonia
   a          Teflon  filters  were  first  wetted  with  200-ul   HPLC-grade
       and  then extracted using 20 mL HC104 solution (5 x 10'5 M) from a
      batch.   The Nylon filters  were  extracted with  20-mL  DI-H20.    A
         reciprocating shaker was  used to agitate extraction vessels  for
  -  ^e Na2C03-coated  tube extraction  solutions  were analysed for  anions
    N°2"»  N03",  S03'(  SC-41");  the citric acid-coated  tube extracts were
       for NH^"*".  Anions were analyzed using ion chromatography  (Dionex) ;
       determinations  were done  with  a  colorometric method  (Indophenol)
        &utoanalyzer.    Anions  were determined for  both Teflon and Nylon
        Ammonium and acidity were  analyzed for the Teflon  filter.

      dlty determinations were made  using an acid addition method with a
       and micro combination electrode.  Duplicate 1 mL aliquots for each
  a  Were loaded into  small  vials;  a KC1 spike was added to give 0.01 M
  ta uniform ionic strength correction.   Standards were prepared from the
  sCtion solution and certified H2S04, and loaded into similar vials with
 ^P^es-   The pH readings  for  each sample pair were  measured,  using a
Sp0 Vial  of  extraction solution between  each  pair  as  a  rinse.   The
 * s*!  °f the  HC104 matrix was to  provide a  5 x  10'5 M (pH A.3)  baseline
t ta°    on acidity.   This forced  the chemistry  of the  samples out of the
 *n6e  dominated by  weak acids  (pH 5-7) which  would have  been  inter-
       at  the  lower   levels  of  aerosol  acidity.   Based on 24  m3  per
  tiee run,  analytical  uncertainties were <2 neq m"3 (0.05 ppb) for anions,
   1  tt*3  (0.1  ppb)  for ammonium and <5 neq m'3 for aerosol acidity.
                                   693
-------
RESULTS AND DISCUSSION

                                                                    t- 1987-
     The  samplers  were  operated for  a  two-week period,  11-25 August *•  f
The schedules of sample intervals and sampler  service were  designed to
(a) the reproducibility of  the sampler  and  (b)  the  effect of  delayed
changing, i.e. letting the  unexposed  or exposed samples  remain in the
after setup or before pickup.   This consideration was addressed  becaus
the samplers in remote areas might necessitate  such  delays.
                                                                     rf ffltf6
     The results for gaseous and aerosol concentrations  are shown in    f0t
I.   Unrun/field  loaded  samples  gave  excellent blank  values,  shown
August 11 and 17.   The  sampling interval for August 18  was subdivided  ^
twp 12 -h  runs for  unit B2 .   On this  date,  pollutant  concentrations
higher during  the  daytime  by  15  to  250% for  gaseous  and aerosol sp«c  e
For the side-by-side comparison, the  two 12-h data were  combined  to coUP
with the values of unit Bl .
     Gaseous  concentrations were  determined by the  difference of at
concentrations  between the  first  (X) and  second (Y)  denuder  tubes.    fl
ratio of  values (Y/X)  gives  an indication  of  the NOX artifact cc
for  each  specie;  for  HN02 and  HN03 ,  these were  0.18+0.15 and °-
respectively  (n-20).   The large range for  HN02 seems to contribute
poor precision  found for  this gas.
                                                                      -
     With  the  exception of HN02 ,  the  reproducibility between  samplers
excellent  for  the gaseous  species (Figure  2a  &  2b) .   For aerosol c0i
ents, the  agreement was good,  but not excellent (Figure 2c & 2d) , and  ioti
was  probably  associated with analytical  problems.   The denuder extraj  o*
solutions were  analyzed within  1 week;  the  aerosol filters were store »
several months  before  extraction.   This  may have contributed to i
dissolution of  aerosol  species.
                                                                       Tl&t
     The  results  for  matched  sample  pairs  (simultaneous  pickup)   ^O
compared  to  sample pairs  in  which the pickup  was delayed for  24-48 *> ^
one  sample.    The relative  differences   as  percent   (mean+std.)  were  ^
culated for  sample pairs, along with  the absolute differences  (Tabl6 '  s
The  low  concentrations for many  samples  caused small absolute differ
enc
sma1
to be  calculated as high  relative differences.   Although this is
dataset, this analysis  demonstrated no bias caused by the delay in P

     The  amount  of  N03~  which was  recovered  from  the  Nylon filtei: ^Oi
relatively  high:  Nylon/(Nylon+TefIon)  -  0.28+0.13   (n-20).    It  is  Oi
possible to know whether this was  caused by simultaneous volatilizat;5.o  )
NH3 and HN03  (i.e.,  no  net loss of aerosol acidity from  the Teflon f1   0{
or HN03 alone,  driven   off  by acidic  aerosol  and,  hence,  a  net l°s .^t
acidity.  The magnitude of this "break-through" means that the upper  1  cf
of acidity  loss  is  relatively high.   However,  the high concentrati011  Of
ambient NH3  and  low HN03 are  consistent with  the very  low occurred  u
aerosol  acidity,  thus   it  is  unlikely  that  this  artifact was  a pr°
during these measurements.
                                    694
-------
idg  . °   to  excellent reproducibility was found in replicate operation  of
    icai  systems.    No  differences  were  noted for  samples  which were
       immediately compared to  samples  that remained  in the field  after
        for  24 to  48 hours.   Apparently,  the  denuders served to  isolate
    filters,  and  diffusion  through  the  impactor  to  the  denuders   is
    °iently  slow  to prevent post-run  contamination.


      1DGMENTS

lalj  We  gratefully  acknowledge   to  the U.S.  EPA Health  Effects Research
c0j,rat°ry  and  the Dutch  RIVM  for  funding  and mutual  support  of this
ti,S   rative research.  Also,  we  are  grateful to Dr.  Robert  Stevens  of
t0 j  ^PA  for  suggesting and encouraging the use  of  this new technology and
        Stone  of  URG  for his skillful and  rapid construction of  the  ADS.
      thanks  to  Bernard van Elzakker, Gerard van der Hooff and Jos.  Neele
      for field  and  laboratory  assistance.
    ^•S.  Environmental Protection  Agency,  1987.  An Acid Aerosols  Issue
    Paper;  Health Effects  and Aerometrics.   Research  Triangle Park, NC,
    EPA Report ECAO-R-0140.

    Possanzini, M. , Febo, A. and Liberti, A.,  1983.   New design of a  high-
    Performance   denuder  for   the  sampling   of atmospheric   pollutants.
    4tmoS. Environ. II, 2605-2610.
           W.A.H.  and  Diederen,   H.S.M.A.   (eds.),  1987.    Ammonia and
   Acidification. Proceedings of the EURASAP Symposium, 13-15 April  1987,
   Bilthoven, The Netherlands, available from RIVM, 327 pp.

   Dossier,  T.  Stevens,  R.K.  and Baumgardner,  R.E.,  1987.   A study of
   the performance  of annular denuders and preseparators , in Proceedings
   S£_the  1987  EPA/APCA  Symposium on Measurements of  Toxic and Related
   4lr_ Pollutants. EPA Report 600/9-87/010, pp 168-173.

   Koutrakis, P., Wolfson,  J.M.,  Slater,  J.L,  Brauer,  M,  Spengler, J.D.,
   Stevens,  R.K. ,   and Stone,  C.L.,  1988.    Evaluation  of  an annular
   ^enuder/filter  pack  system  to  collect  acidic  aerosols  and  gases.
   Submitted to Environ.  Sci. & Tech.
                                  695
-------
Table I.    Differences  for   side-by-side   sampler   results
            RIVM/UMDNJ/USEPA Bilthoven Comparison - Summer 1987.
        a.  GASES
MATCHED3

NH3
S02
HN03
HN02
b . AEROSOLS
D
4
4
4
3

%RDb
0±4
3±4
13±16
9±35

DELAYED
Dc
0.2
0.06
0.05
0.4

MATCHED

H+
NH4+
S04~
N03~
n
3d
3d
3d
4
%RD
-20+34
20±18
5+8
1+10
D
1.6
94
34
13
n
3
3
3
2*

%RD
2±1
-2±8
1+11
-9±24

D
0.3
0.3
0.05
0.3

DELAYED
n
2e
2e
2e
3
%RD
1+1
1±1
2±3
3±4
2
0.8
3
2
6
a.

b.
c.
d.
e.
Matched - pickup immediately following sampling  intervals.
Delayed - field pickup was delayed (24-48 h)  for second sampler.
%RD - 2*(X1-X2)/(X1+X2)*100;  Mean + std.dev. are given.
                                                           Q
D — (X^-X2);  Means are given (gases — ppb;  aerosols — neq m~°).
1 outlier removed.
1 filter lost.
-------
 ppb
                   OASES
10
                 AUQUBT18S7

              0-Sta1   X
                               »
ppb
                 OASES
20
10

5

a>
to

600
IfMl

3OO
0
NH3
> . * . > -
802 s \ •
_ t k* 	 k 	
' ' *
O-8Ha1 X-Sb2
AEROSOLS
necj TO J
NH4*
1 * 	 * 	
*
* • k »
k
804-
. k i »
2
1
0
1
K

40
on

0
1OO
n
HN03
k »
" * * k " > i
HN02
... *
k » »
• k k « »
« *k K k
• - • *
I W 89
AUGUST 1QS7
O-Bita1 X-Sib2
AEROSOLS
neq m
H*
«
* ' •
• » E
-------
                S02(q)
                                 ppb
                                                            u
                                                            11
                                                            14-
                                                            U
                                                             I
                                                            1J
                                                            U
                                                            V4
                                                            U-
                                                             I
                                                                             HN02(g)
                                                                             "in on)
(c)
Particle Acidity
                           nanoiq/i
             a
            -U-T-
                                                            (d)
                                                                           Told N03-
                                                                                          nan
                                                                       w"

Figure 2.  Conp*rlson* of  aluultaneoua samplers for concentrations of gases (a) SO, and  (b)
           and serosols  (c) acidity and (d)  nitrate.
                                             698
-------
           OF A  SAMPLING PROCEDURE FOR  LARGE
           PARTICLES:  PRELIMINARY RESULTS
      D.  Lane,

        .  Randtke,
     E-  Baxter,
Department of Civil
Engineering
University of Kansas
Lawrence,  KS  660^5

Department of Civil
Engineering
University of Kansas
Lawrence,  KS  66045

Department of Civil
Engineering
University of Kansas
Lawrence,  KS  660^5
"Uh .G°Urate quantification of the individual  chemical  species associated
8y8j.  py deposition requires a highly specialized  sampling system.   One  such
^Paot-" Currently  under  development  consists of a teflon-coated  glass
U u  c  J-n series  with denuder tubes and a filter pack.  The glass impactor
lUamo®CJ to Prevent  large particles,  i.e.,  those having  an  aerodynamic
         (D   )  greater  than 2.5 microns (urn),  from  entering the first
        However,  the sampling and  collection efficiency  of the glass
     Or for large particles has not been well  characterized,  and  questions
 .y   regarding particle  bounce,  mass  loading  limitations,  the use of oil
^rtip iease coatings  to  improve  collection efficiency, chemical  artifacts
%j  '  in the inlet  {particularly on  the impactor surface), etc.   The
%ipl.ive of the on-going research project described herein is to  develop a
4t>aivln.S Procedure  that  will  permit accurate gravimetric and  chemical
^tjf    3  of the  large  aerosol  fraction without formation  of  chemical
feht '  -ts or interference with the sampling and  analysis of gasses  and fine
    -les.

BQ   ased upon  the  preliminary results presented herein:  1) the actual cut
aam,M  of a 4.0-mm  glass  impactor was  found  to  be 2.64 urn;  2)  isoaxial
    nS win  be  necessary to accurately  sample  large particles,  mandating a
     '*">&!  inlet oriented in the direction of  the oncoming  wind;  3) oil is
  j    to grease  as an impaction surface  coating, but  requires  a  vertical
  ,  Or  orientation; 4)  a  modified  impactor  design was developed to
    tate gravimetric and  chemical  analysis  of impacted particles;  5)
     lcant deposition of  8-pm  particles occurred in the  first  bent-tube
      tested;  6)  several  oils  were  determined  to be compatible with
  ,   "-nation of  trace levels of anions  by  ion chromatography;  and 7)
   °ne  oil  did not interfere  with  extraction of sulfate  from large
                                 699
-------
INTRODUCTION

     Despite the major public attention that has been focused on  the P ^at
of "acid rain", it has been recognized  by scientists for some time  now   f
a substantial  fraction of the acid is  actually deposited during  dry
due to the interaction of gasses and aerosols  with plant surfaces
ground.   So that reliable  estimates of the loads of nitrogen and
associated with this dry deposition can be made, scientists  at the  u. '
and elsewhere  have been working to develop quantitative methods to    jp
and analyze the major nitrogen- and sulfur-containing species  pi"e  Q^e<
gaseous  and aerosol form  in ambient air.  These include nitrous   i
-------
!)v ea°h  of  these devices, including:  1) particle resuspension or  bounce;
ijpi. atilization of constituents of interest  from the surfaces of large
* "it°^e9 '  and -^  reactions of gaseous  constituents with trapped particles
     any grease or oil used to trap  the particles.  Furthermore, neither
     la  well suited to  gravimetric and  chemical  analysis of the large
     e fraction, which  would be highly  desirable in view of  recent
  ,6 that  significant amounts of nitrate and ammonium may be present in
   ap8e-particle fraction.

Sli   obJective of  this on-going research investigation is to develop a
!([e In8 procedure for large  aerosols that  will:  1)  permit accurate
*«aJ?lnation of the mass and chemical composition of large particles
fa j  ln  ambient air over a 2H~hour period; 2) employ a device that can be
'"tepf  conjunction with  an  annular  denuder  and filter pack without any
Vu6r>ence  in the determination of  the gaseous  and fine-particle
Sgh  nts; ^ avoid the formation of chemical artifacts; and 4)  be simple
\^ to deploy in a nationwide monitoring network.  This paper describes
^D  Oach belng taken to develop the sampling system and presents  some of
            results of the research.
'1,  he investigation began with a  careful assessment of the  advantages and
\ajaritages  of various  devices able to  collect large  particles and
 leflle for use in conjunction with the annular-denuder filter-pack system.
           ed  Elass impactor,  modified to  facilitate gravimetric and
      analysis of the particles  captured on  the impaction surface, was
,Vy  ed as the best available alternative.  This device is very similar to
''fl0n °n~ coated glass impactor used by Stevens  et al. , 9  but  a removable
Vfan disk mounted on the end of  the denuder tube is used  as the impaction
5ye» as shown in Figure 1.   This arrangement was conceived and developed
     authors  (in cooperation with University Research Glassware) as part
       project previously funded by the U.S. EPA (Grant  R81 2280-01-0) .

SaJ16 ""^movable Teflon impaction disk  (Figure  1)  can be  coated with
%,  '  covered  with an oil-saturated filter,  or  fitted with  an  oil-
 1 
-------
                                                                   c USe
24-hour sampling period.  Thus,  regardless of which coating material i
to reduce particle bounce,  there will be effects on the perform3"1
operation of  the impactor sampling system that must be considered.
                                                                     a
     To gain  insight into certain aspects of the impactor *s performan°e
operational  characteristics, experiments were conducted to examine:  '
cut-point of the 4.0-mm teflon-coated glass impactor;  2)  the seve
                                                                 r! J
                                                                   a
problems  associated with nonisoaxial sampling;  3)  inlet losses;
particle loading limitations.   In addition, a variety  of oils and Sr  ^
that might be used to prevent  particle  bounce were examined, and preli"1  ..
experiments were conducted on  the most  promising  candidates to deter1  ^
1) whether they  would be compatible  with analysis of  trace anions  W  ;
chromatography; 2) whether they would permit accurate gravimetric an  L $
and 3)  whether  they would react with  gaseous constituents  to form
artifacts. Future experiments will  explore these issues  in greater
and examine the interactions of gaseous constituents and  large p

RESULTS AND DISCUSSION

                            Impactor Cut-Point
                                                                   lJ8t
     A single- jet, teflon-coated glass  inertial impactor, with a nofflina  ^i
diameter,  W,  of  4.0 mm, was  obtained from University  Research Gla"  ^t
Carrboro, North Carolina.   The  impactor  was  designed  usin^  $
recommendations  of Marple and Rubow.10  The actual nozzle  jet diamet6  ^
measured at 4.0 mm.  The jet-to-plate distance, S,  was measured to b6g/tli
mm, resulting in an actual jet-to-plate distance to jet diameter rati°'
of 1.5.
                                                               •  a a
     Laboratory testing of the glass  impactor was  conducted using •  $
tunnel designed to produce low turbulence and isokinetic  flow conditi0  ,i
the impactor  inlet.   Flow through  the impactor was accomplished u9  $
vacuum pump.   For  all  tests,  the  flow rate through  the impact0  $
regulated at  16.7  liters  per  minute (Lpm) using a Brooks  flowmete  jii
needle valve.   The  test particles  were generated  using a  Berglurl ^
Monodisperse  Aerosol  Generator  equipped  with  a  Krypton-85 Q ^
neutralizer.  The size and shape uniformity of the generated aeros°-L ^
verified using a  scanning electron  microscope.   Particles  with aerod*
diameters of  1 . 1  ym,  2.0 ym,  2.5  um, 4.3 ym,  and  6.3  pm were
                                                                   .
during the  4.0 mm  impactor  tests.   A  complete  description °
experimental  procedures is given by Baxter etal.11
     The  experimental collection efficiency curve for the  4.0 mm impac
                                                         .
shown in Figure 2.   The particle  size is .expressed  on  the abscis
dimenaionless  form as the  square root of the Stokes  number, St.  The  ^
in the figure displays a sharp cut point and is S-ahaped.   Table 1  S^v' ^
experimental parameters for  the impactor tests as well as the aeroo/  .j.
cut diameter corresponding to the experimental and theoretical St50 va,c^
The equations  governing the  relationship between  St and aerodynafl
diameter  are given elsewhere.10 The experimentally determined aer
particle  cut  diameter for the  4.0-mm glass impactor was 2.64 um, ifl
agreement  with the 2.50 ym predicted from theory.

                           Nonisoaxial Sampling

     Two tests were  conducted using ammonium  sulfate  particle^  f
aerodynamic diameters of 8.0 ym.  In each test, two 4.0-mm straight"
impactors  fitted with backup  filters  were placed  in the  wind tunn^
                                  702
-------
     i at various angles from  the wind direction.   Also placed in the wind
an e  during each test were two sampling filters operated under isoaxial
l^  is°kinetic  flow conditions to accurately determine the total particle
(t a •  The sampling angles of the impactors during  the  first test were  0°
te °a*Ul) and 45°.  The sampling angles of the impactors  during the second
  1 were 80° and 90°.
       fractional  collection results  for  these tests are presented in
%(jo  2.  The data show that  as  the sampling angle  increases away from an
f0p xial condition there is  a corresponding decrease in sampling efficiency
90,o '°-um particles.  The drastic decrease in collection efficiency for the
IW^°sition  indicates that elutriation is predominating at the 8.0-ym
Vlg icle size.   This is supported  by  the results from  several  flow
jenualization  tests that  were  conducted using  a simple smoke-wire to
8v rate  fiow streamlines. The  flow visualization test in Figure 3 clearly
S   the sma11  regional sampling  influence that is  available when the
°°rir|Gt0r is operated at a  position  90° away from an  isoaxial sampling
the     n-  It  should also  be noted that inlet losses tend to increase as
  Sampling angle increases  away  from isoaxial.

                    Sedimentation and Mass Loading

IQ  T° determine whether 8.0-ym particles would settle in the horizontally
W.  d  glass-impactor inlet  and  to determine the bounce and particle-
Hgf, n8 characteristics of the system using silicone vacuum grease,, tests
*opri3lng tnafc significant  bounce occurred at loadings of 31 and 48 yg.
»tijB0u8h  the effects of gravity  and specific particle load limitations for
<*vt   Particle  sizes will  be  different than what was observed here, it is
  °^s that these problems can  not be overlooked.

W Bother set of  tests was  conducted with the impactor, denuder, and
th$Up fUter  in  a vertical sampling position while using an oil coating on
(Han ilnPaction  surface.   A Teflon-coated 90°  bent-tube inlet section
Of tufactured  by  University Research Glassware) was slipped over the outside
!Whe Impactor inlet,  permitting isoaxial and isokinetic sampling.  The
*H
-------
                                                                 * ft
tunnel.  Sodium fluoresceln  particles with aerodynamic diameters ol °  .
were  generated; and  the  run time  for  the second  test (720 minute
double that of the  first (360 minutes) so that the particle  loads of
tests  would be significantly different.   Table 4  presents  the
collection results  for  these tests.
     The data show that  there is a significant amount  of  particle dep°  ^
occurring in the bend.   Pui et al.'2 have suggested that  particle dep"   fl[
efficiency in bends should approach zero for the finite Stokes numb
0.1 at Re ~ 1000 and 0.2 at Re = 100.  The flow Reynolds  number of th1
bend  used in these tests was 2120 and the particle Stokes number was  ^
Therefore, particle deposition under these  conditions should havei
minimal.  The deposition which did occur, however,  can be attributed t ^
the presence of an abrupt transition from the bend  into the straign^_
section of the impactor (particles were visible on the  lip of the
inlet);  and 2) the  blunt end of the  bent-tube inlet,  which undou0""
increased turbulence in  the bent tube.

                                                                    X
     To reduce deposition in the bent-tube inlet, it was  redesigned a $
-------
   "r° ^oughly assess the weight stability of the oils, a small amount  (~ 2
  °f 6ach °^ was Placed in fcne b°ttom  of three large (9-cm) plastic petri
fiav93'  The  dishes were then loosely covered and placed in a balance  room
      good  temperature and  humidity  control.   The  results for  the
      rand  19 and sllicone olls are  ah°wn in Figure 4.  The weight  changes
   generally less than 200 yg/gm.  Such a change is negligible «  10 ug)
   t!le small  amount « 0.05 gm) of  oil needed to saturate a 1 3-mm  filter,
  n   potentially significant for the larger  quant ities 'of  oil needed to
 a rate a porous glass disk.  However, it is quite possible that the weight
V 8Ss ot)served were associated with  the petri dishes rather than the oils.
"Isk  a°Qurat,e  tests using  teflon impaction disks fitted with porous glass
  3 are planned.

I6a To determine whether various oils and greases contain anions that might
Of 01 int° solution during extraction  of  the collected particles, solutions
    3 and greases dissolved in pentane  were extracted with ultrapure water
        in sequential and batch tests.  The  aqueous extracts  were then
 a      U3ing ion chromatography.   The results are shown in Table  5.  The
 rles for sulfate  reflect the  presence  of  small quantities  of
   actable sulfate in the silicone grease and vacuum-pump oils.

      determine whether silicone oil would  grossly interfere with the
         of sulfate from collected particles, 8-ym  ammonium-sulfate
        generated in  the wind tunnel  were  sampled isoaxially  and
%d   tlcally and collected  on:   1)  oiled and unoiled Teflon filters; 2)
81«as and unolled polycarbonate (Nucleopore) filters; and 3) an oiled porous
^fat.lmpaction surface.  Collection,  determined on the basis  of extractable
fUtte. was 8.3* less on the oiled Teflon filter than the unoiled Teflon
   ri  1 0£ greater on  the oiled polycarbonate filter than  on the  unoiled
   ar Donate  filter; and 20% less on the oiled  porous glass  disk  than on
        d Teflon filter.   Since the sampling  accuracy of the test was
        to  be ± 10-20$,  the results only prove that the  oil  did not
ftir. ^Jy interfere with extraction of sulfate.  The low collection  determined
\l   oiled  porous glass impaction disk indicates that further testing for
    interference by the porous disk is warranted.
        have also been conducted to determine whether gaseous  pollutants
       to  react with Teflon impaction  disks or with oiled filters  placed
     of them.  The experimental apparatus  is  shown in Figure  5,  and the
      of  two  sets of tests are presented in Tables 7 and 8.   Nitric acid
      react with plain Teflon impaction  disks, but it did react to a very
     ^Perhaps negligible) extent with oiled Mlllipore filters.   Additional
    are in progress.
                                  705
-------
CONCLUSIONS
vail*"!!
 1.   The  Teflon-coated glass  impactor appears  to  be  the best
     device  for sampling  large  (OAE  > 2.5  ym) ambient
     simultaneously with gasses  and fine particles;  a modified iinPaicai
     design  has been developed to facilitate  gravimetric and chem
     analysis  of large particles.

 2.   The  cut  point of a iJ.O-mm glass impactor was experimentally Ae^e^^.
     to be  2.64 ym, in close agreement with the predicted  value of 2.50

 3.   Nonisoaxial sampling  of 8-ym particles resulted in a substan ^
     reduction in sampling  efficiency.  Hence,  accurate sampling
     field will require  a  vane-mounted system or other  means of
     the  sampling inlet into the  wind.

 14.   Sedimentation of  8.3^-ym particles was observed  in  a  horizon
     mounted sampling system.
                                                                   y f 3^
 5.   Significant particle bounce  was detected on a greased impaction su  .  a
     at a total particle  mass loading  of  38 yg, representing  70> ^ Of
     monolayer for the  3.7-mm-diameter  area actually  impacted.    d^g
     grease will limit  mass  loading to  very  low levels, predug0f
     gravimetric analysis, unless a  rotating surface  or  other  n>ea
     reducing  areal loading is provided.
                                                                     f\d
 6.   An oiled porous glass impaction surface gave 98?  collection eff ic ^%
     of 8-ym particles at an impacted mass loading of 79 yg, deruonstr* ^
     the  superiority of oil over  grease.  Use of oil necessitates a ver   ^e
     impactor  orientation,  so a bent-tube  inlet  or  other means wil
     required  to achieve isoaxial sampling conditions.
                                                                    let3
 7.   Significant particle losses occurred  in  the  first bent-tube  in
     tested; modified inlets have been  designed and constructed.
                                                                      *
     Vaseline  was determined to contain small amounts of extractable
     that could  be removed by extraction with pentane and water.  A s    ef
     grease was  found to contain significant amounts of sulfate, even  3
     repeated extractions.
                                                                    tab16
 9.  Vacuum pump oils and a silicone  oil were found to be reasonably s flSl
     gravimetrically,  to contain negligible amounts of extractable an ^
     and to be compatible with analysis of  trace  levels of anions by
     chromatography.
                                                               i  nf 8'^
10.  Silicone oil did not interfere with the extraction and  analysis 01
     ammonium-sulfate particles collected on either  Teflon  or  polycarb
     filters. Recovery from an oiled porous glass disk was 20% lower
     the control samples; but  since the  difference was not  statisti
     significant under the conditions  of the  test,  further  study
     recommended.

                                                                  /«' ^
11.  Teflon disks exposed to nitric acid concentrations of 6-25 vS/m tei/
     not form any  extractable artifacts in a one-hour test.   Approxi"1 •
     0.3 yg of extractable nitrate was found  on  silicone-oiled Mil1 iP
     filters exposed to 91-702 yg/m3 of nitric acid,  suggesting
     artifact formation will be negligible under ambient conditions.


                                   706
-------
 CIO|QWLEDGEMENTS

 Jai TTlle autnors wish to express their appreciation  to Gary Guinn and Mehdi
 tQ [   f°r  assistance in the collection of chemical  and gravimetric data and
 the Srr^ Stone  and Ed Clark for expertly manufacturing various components of
 0)  SampHng systems.  This research was financed by a grant (CR-814613-01-
 pQ, r°ro the U.S. EPA.  The contents do not necessarily reflect the views and
 tan °^es °f the EPA, nor does mention of trade names or commercial products
       -e  endorsement or recommendation for  use.
 1.  ,
    APpel, B.R.,  e_t_al_._, "Artifact  Particulate  Sulphate  and  Nitrate
    Formation on Filter Media," Atmos.  Environ., 18(2), 409 (1984).

    Forrest, J., et al., "Determination of  Atmospheric Nitrate and Nitric
          Employing a Diffusion Denuder with a Filter  Pack,"
          !., 16(6), 1473 (1982).

    Anlauf, K.G.,  H.A.  Weibe, and P.  Fellin, "Characterization of Several
    Iritegrative Sampling Methods for  Nitric Acid, Sulphur  Dioxide and
    Atmospheric Particles," JAPCA,  3j),  715-723 (1986).

    p03sanzini, M.,  A.  Febo, and  A.  Liberti, "New  Design  of  a High-
    performance Denuder for the Sampling of Atmospheric Pollutants," Atmos.
    SSllron.., 17(12), 2605-2610 (1983).

    perm,  M.,  "A  Method  for Determination of Atmospheric Ammonia," AtmosL
    Savlron.. 13,  1385-1393 (1979).

    Durham, J.E.,  e^_al^, "A Transition-Flow Reactor Tube  for Measuring
    Tracer  Gas Concentrations," JAPCA,  36,  1228-1232 (1986).

    Appel,  B.R. , et al., "Simultaneous  Nitric Acid, Particulate Nitrate and
    Acidity Measurements in Ambient  Air," Atmos. Environ., 1_4,  549  (1980).
 0
    Sickles, J.E., et al., "Performance and Results of the Annular Denuder
    System  in the Sampling and Analysis of  Ambient Air Near Los Angeles,"
    proc.  EPA/APCA Symposium on Measurement  of Toxic Air Pollutants (1986).
 9
    Stevens, R.K., T.L.  Vossler,  and R.J.'Paur,  "Evaluation  of Improved
    inlets and  Annular Denuder Systems  to  Measure Inorganic  Air
    p°llutants," paper submitted for publication (1987).
'0
          i, V.A., and K.L. Rubow,  "Theory  and Design Guidelines,"  Ch. 4 in
         e Impactor Sampling and Data  Analysis, edited by J.P.  Lodge,  Jr.
       T.L. Chan, Am. Ind. Hyg. Assoc.,  1986.
1]
    Baxter, T.E., D.D. Lane,  and S.J. Randtke, "Initial Performance Testing
    of  a Glass Jet  Impactor Designed for  Use  in Dry Acid  Deposition
    Sampling,"  Proc. Amer. Assn. for Aer. Res., Seattle,  WA,  1987.

    Pui,  D.Y.H.,  F.  Romay-Hovas, and B.Y.H. Liu, "Experimental  Study of
    particle Deposition  in Bends of  Circular Cross Section,"  Aerosol
    Science and Technology, 7(3), 301 (1987).
                                  707
-------
13-   Moss, O.R.,  and J.L.  Kenoyer, "Use and Misuse:   Operating Guide," Ch. ?
     in Cascade  Impact or Sampling and.J3ata Analysis, edited by J.P. Lodge,
     Jr. and T.L.  Chan," Am. Ind. Hyg. Assoc.,  1986.
Table 1.  Experimental Parameters and Comparison of Actual and Theoretical
          50% Aerodynamic Cut Diameters for the 4-mm Impactor.

                    Actual jet diameter            4.0

                    S/W                            1.5

                    Rp € 16.7 LPM               5670
                     LJ


                    /St50   Experimental           0.471
                            Theoretical            0.49

                    50% D  ,  pm   Experimental     2.64
                         A.U
                                  Theoretical      2.59

                    50% DAC, Discrepency, %         4
Table 2.  Fractional Collection Results for Nonisoaxial Sampling Tests with
          8.0-ym Aerodynamic-Diameter Particles

                                  FRACTIONAL AMOUNT COLLECTED, %
SAMPLING
ANGLE
0° l
45°
80°
90°
IMPACTOR
INLET LOSS
3-0
7.5
73-6
45.0
BACKUP FILTER
CAPTURE
97.0
92.5
26.4
55.0
RELATIVE COLLECTION
OF TOTAL POSSIBLE2
99.1
90.5
28.7
8.5
     1 0° = Isoaxial
     z Based on total mass collected on isoaxial, isokinetic filters.
                                    708
-------
Table 3.   Fractional Collection Efficiencies of Impactor Sampling System Under  Increasing
         Particle Loading Conditions (Isoaxial  Sampling,  Horizontal Orientation,
         Aerodynamic Particle Size - 8.34 urn;  Silicone Grease Coating.)

                       	FRACTIONAL COLLECTION.%
 TEST  RUN TIME,   INLET  IMPACTED  GASKET  DENUER  BACKUP FILTER  TOTAL MASS   IMPACTED
 _Nq.    MINS.     LOSS             LOSS    LOSS      CAPTURE      LOAD^ jig    MASS, y&
          20
          20
          33
         120
         270
         360
8.8
7.4
8.1
1.8
1.9
2.0
90
91
91
93
82.0
84.9
0.2
0.2
0.7
3.1
1.7
0.0
1.2
4.4
3.2
3.1
8.6
8.2
2.2879
3.0484
4.4209
13-5952
37.5978
56.9645
2.0599
2.8025
4.0492
12.6678
30.8390
48.3842
Table 4.  Fractional Collection Results for Impactor Tests  Using Bent-Tube Inlets and
        Silicone Oil Coating (Vertical Sampling Orientation;  Aerodynamic Particle
        Diameter =8.34 ym)

                          FRACTIONAL COLLECTION AMOUNT, %

TEST
Ho.
1
1
2
2

IMPACTOR

A
B
A
B

BEND
LOSS
6.9
11.2
6.5
11.0
IMPACTOR
INLET
LOSS
0.3
0.1
0.6
1.9


IMPACTED
91.4
87.0
92.4
86.0

GASKET
LOSS
0.1
0.4
0.1
0.2

DENUDER
LOSS
0.4
0.2
0.1
0.3

BACKUP
FILTER
0.9
1.0
0.3
0.6
TOTAL
MASS LOAD
pg
59.1873
61.2770
85.5920
91.3278
IMPACTED
MASS LOAD
Mg
54.0796
53.3074
79.0904
78.5850
Fractional amount of impacted particle load collected  by  irapactors after correcting for
aaaa collected in bend.
                       TEST No.  1-A:   Impacted Fraction =  98.1*
                       TEST No.
                       TEST No.
             1-B:
             2-A:
                       TEST No.  2-B:
            Impacted Fraction - 98.0$
            Impacted Fraction = 98.6%
            Impacted Fraction » 96.7%
                                       709
-------
Table 5.  Impurities Found in  the Aqueous  Extracts of  Pentane Solu
          Containing Various  Oils  and Greases1
     Oil or Grease
       Examined
Vaseline
Silicone Grease
                     1
                     2
                     3
                     1
                     3
                     4
                         Type  of
                          Test2
                   NO,-N
                    1.6
                    2.6
                    1 .0
                    3.0
                    2.4
                    1.0
                    1 .6
                           Concentration,
                      P0.4.-P

                        ND
                        ND
                        ND
                        ND

                        ND
                        ND
                        ND
                  N03-N
                                                                       tfD
Duo-Seal Oil
                     1
                     2
                     3
                     4
                    0.2
                    0.4
                    0.2
                    0.2
                        ND
                        ND
                        ND
                        ND
Silicone Oil
Fisherbrand 19  Oil
Silicone Oil3
Silicone Oil3
Silicone Oil3
Fisherbrand
Fisherbrand
Fisherbrand
            19  Oil3
            19  Oil3
            19  Oil3
1
2
3
4

1
2
3
4

1
2
3

1
2
3
B
B
B

B
B
B
                                        ND
                                        ND
                                        ND
                                        ND

                                        T"
                                        ND
                                        ND
                                        ND
.8
,7
,6

,2
,2
,2
ND
ND
ND
ND

ND
ND
ND
ND

ND
ND
ND

ND
ND
ND
   Greases were  dissolved to  156 in pentane, oils to '\Q% in pentane;
   pentane solution  were extracted with 20 mL Milli-Q water,  which
   analyzed by  ion chromatography.

   S = sequential  (aqueous phase  removed entirely after each
   Milli-Q  water and a fresh 20  mL of water added for the next  extrac
                                                                       t&
   B - batch (3 aliquots of oil  in pentane were
   separately;  the reported results are averages of
                                                   extracted  and
                                                   3 samples).
                                                                    t  g
   Exposed to  indoor  (laboratory) air  for nearly  2  weeks during a
   stability study.

   T =  trace
                                                                    *
                                   710
-------
      Percentage Recovery of Selected Anions  from  400-pg/L Standards
      Extracted with Pentane Containing Oil  or  Grease1

      Oil or Grease            	Percentage Recovery	
 	Examined2	        M02-N      PO^-P     M03-N       SO..

 Duo-Seal Oil (6)                -         -       101±4      107±3

 Pisherbrand 19 Oil (4)        106±7      104±8     104±6      115±7

 Celine (5)                  102±1       99±1     103±1      100+5

 Sllioone Grease (5)            103±3      103±5     104±5     109±12

gt6aaes  were 1% solutions  and oils were  10% solutions; 20-mL portions of
 afidards were extracted with 10-mL portions of  pentane solution.

 ™ber of  replicates  shown in parentheses; vaseline and silicone grease
         were "precleaned"  by extraction  with ultrapure water.
      Artifact Formation on Teflon Impaction Disks Exposed to HN03 (One
      hour  test at  16.7 Lpm)

     Sample             HN03-N in air, yg/m3          N03-N on disk,  ug

         Blank                   0.00                         0.05

   stern Blank 1                  0.26                         0.08
      rn Blank 2                  0.20                         0.08
      rn Blank 3                  0.16                         0.07

      Run No. 1                 24.7                          0.12
      Run No. 2                  7.32                         0.06
      Run No. 3                  6.17           -              0.08
     Artifact Formation on Teflon Impaction  Disks Holding Millipore
     Filters Saturated with Silicone Oil (One hour test at  16.7  Lpm)

                    HN03-N in air, yg/m3       Nitrate-N on disk,  ug
 Rea
   gent Blank                0.00                       0.35
       Blank 1               0.45                       0.31
  stern Blank 2               0.38                       0.28
      n Blank 3               0.32                       0.30
     Run No. 1             702                          0.61
     Run No. 2             163                          0.64
 rest Run No. 3              91 .0                        0.56
                               711
-------
                                         Teflon-Coated
                                         Bent-Tube Inlet
                            •Teflon Sleeve Coupler
                           Teflon-Coated
                           Glass-lmpactor
                           Nozzle (4.0 mm)
                           Teflon Spacer

                          •Teflon Impaction Disk


                           Annular  Denuder with Pedestal
Figure 1.  Modified Teflon-Coated Glass Impactor with Bent-Tube

                             712
-------
   100
 o

 S
 o


    50
 u
 UJ
 o
 o
           'St50 = 0.495


          50% DAE = 2.64 nm (2% Discrepency)
     0.2
0.4
0.6
                                          0.8
                                     1.0
                                                                   1.2
      Experimental Collection Efficiency Curve for a JJ.O-mm Impactor
3.
    Plow Streamlines for a 4.0-mm Glass Impactor  Operating  at 16.7

        with a Wind Speed of 800 fpm
                              713
-------
               300
               200 -
             E
             ra
             o
             o>
             id
             O
             CO
             "5
               -200
                                      A Oil A (1)
                                      D Oil A (2)
                                      O Oil A (3)
                                                      A Oil C (1)
                                                      • Oil C (2)
                                                      • Oil C (3)
                                                  10
                                         Day
                                                                 15
Figure J».   Gravimetric Stability of Vacuum-Pump Oil  (A) and
            Diffusion-Pump Oil (C)
    AMBIENT
       AIR
                COMPRESSED GAS
                OR SUPPLEMENTAL
                  FEED SOURCE
                        FLOW
                     CONTROLLER
GLASS FIBER
  FILTER
   FLOW
CONTROLLER
GLASS WASHING
 BOTTLE WITH
DIFFUSION TUBE
AMBIENT^
AIR
GLASS FIBER
FILTER
_*
FLOW
CONTROLLER

fe

GLASS IMPACTOR
&
DENUDERTUBE
VACUUM
PUMP


ABSORBENT
FILTERS
   Figure 5.  Experimental Apparatus for Examination of Chemical Art

                                     714
-------
                  EPA'S INDOOR AIR QUALITY TEST HOUSE
                       2. KEROSENE HEATER STUDIES

                           Merrill D. Jackson
                    U.S. Environmental Protection Agency
             Air and Energy Engineering Reseach Laboratory
                    Research Triangle Park, NC 27711

             Russell K.  Clayton,  E.  Eugene Stephenson, Jr.,
                         and William T. Guyton
                           Acurex Corporation
                             P.O. Box 13109
                    Research Triangle Park, NC 27709


  ^  A llas  leased a house  for  conducting indoor air  quality  research in
 lS81Search  Triangle Park, NC,  area.   Research is being conducted  on the
 Q(Ju °nS of  organic  compounds  from  common building materials,  household
 stj   »  personal activities,  and  combustion  sources.   The  results  of
      in EPA's  small chambers and  in  large  (room  size) chambers  are
       with results  obtained in the  test  house.

*1|:h c   test  house  is  a  typical three-bedroom,  two-bath,  single-story
   t awl space,  frame  house  with   natural   gas  heat  and  central  air
  Ut  niT1S«  The house  is  8  years  old  and  has the energy  efficiency
    es of houses built during the early  1980s.

    nvented kerosene  heater  emissions were  sampled  using Tenax-GC  and
    8°r bents.   Organic emission  compounds  were identified by GC/MS,  and
     ComPounds  were  then  quantified. Two types  of   unvented  kerosene
   tl  Were  tested with  three  replicate  runs  conducted  for  each  type.
  i t m°noxide,  carbon dioxide,  and  sulfur dioxide were monitored  during
 
-------
                                   INTRODUCTION

     The Environmental  Protection  Agency's  (EPA's)  indoor  air  quality
research includes a combination of  chamber  testing,  modeling, and test*   ,
a typical house for organic emissions from  building,  household, and c°   i
tion products under normal usage conditions.  Kerosene hooters  were set ^
in the  large  chamber,   and  the  results  were   reported  earlier  ^   i by
single-story frame houso  leased by EPA  for LAQ research  was  duscrtbe  .£
Jackson et  al.    (1).   The  present   study  envolved  sampling  the  °rs  ^
emissions from unvented kerosene heaters under  normal  usage  condid0
that house.

                              DESCRIPTION

                                                                    mer^
     Two unvented  kerosene heaters  were   chosen  from  the  12  comw   ™
heaters screened by White et  al.(2).   Heater were selection ^as based   u,
on the largest  amount   of  parti culate  emitted in the  large chamber    ^j
The heaters selected were  of  the radiant /radiant (R/R  ~nd the conveC  (
radiant (C/R) types.  The heater to be tested was  placed in the den,    »j
 raan          ypes.     e  eaer  o  e  ese  was pace    n    e     ,
 one window was open 5 cm (2 in.) (EPA's interpretation of  the  manufact   ^
 suggestion to  crack  a  window  during  operation).   The  manufacturer   gj
 structions were  followed in adjusting the  flame  height to a specif*6   jot
 during  operation.  The kerosene  used  was ASTM grade 1-K,  from  the  $&   $
 used  In the  large  chamber studies.   Each heater test  consisted of  *8  j0d,
 of burning.  Health  effects (bioassay) required a 48  hour  sampling P  ^
 whereas the  organic  sampling only needed 12 hours of  sampling.  Thef  t V
 three replications of each heator tost. The  bioassay research will   ( *
 presented in this  paper.  The  organic  sample was  always taken at    t'UP
 hours after  starting  the test;  therefore,  it  did  not Include  any  st
 emissions.

     A  complete  set   of  background  samples was collected  before
 each kerosene heater  test  series.
                                                                      f
                                                                     01
     The house  was  aired  out  by  opening  windows  for  a minmiuni
 hours between tests to remove any  residual  organics  In the  house.

                                                                     r
     The sample  collection points  were in the den  where the neat^n
 placed, in the  bedroom  (corner)  diagonally  across  the house,  sn<*
 backyard. All samples were collected at  a height  of  163 cm (64 in.,  st
 height) from the floor.  Since the  baseline  studies Indicated  that
 insignificant stratif ication in the house  under  these conditions,
 sampling height  was   used  (1).  At  each  sampling  point  a  gas  scfli
 directed to  the  continuous monitoring  system for   carbon  monoxide,  000C'
 dioxide, and sulfur dioxide analysis. The volatile organics (b.p«   ~n $'
 were sampled with Tenax  cartidges  at  flow rates of  250 ml/mi n for  .^y
 Four Tenax  cartridge  pairs were   collected  per  test.   The  semi~"v .
 organics (b.p.   100-30Q°C)  were  collected  on  XAD-2  resin traps
 modified medium flow  (114 L/min or 4 cfm) sampler.   The sampling tirae
hours. A Teflon-coated  glass fiber  filter in  front of  the  XAD
 collected the  particulate.  These   filters  were  changed  every  °
                                                                    *»F   a
     Air exchange rates  were determined during  the  tests  by use of  ^ 1^
 a tracer  gas.  A  meteorological  tower  in  the  backyard  provided     $
 temperature,  wind  speed,  and direction.   The temperatures in  the
 corner bedroom were recorded, using thermocouples.
                                   716
-------
 ot
                             RESULTS

  •^he air exchange  rate  for normal weather conditions  was  0.35 ACH (air
    per hour) with  the window  closed;  with one window  open 5 cm,  it was
  ACH.   The concentrations  of  gases   measured  during  the  12 hours  of
    testing are given In  Table 1. There were  significant  increases above
   und concentrations of all gases during the  operation  of  each heater.
ever» the R/R heater had the  larger increase.
                       TABLE 1.  GAS CONCENTRATIONS
                                   ppm
RUN TYPE
Ba<* ground
C/R Heater
Ba<* ground
R/R Heater
*- 	
RUN
NO.
1
1
2
3
2
1
2
3
SO 2
0.00
0.00
NA
0.01
0.00
0.02
0.00
0.02
- 0.00
- 0.02
- 0.02
- 0.00
- 0.04
- 0.01
- 0.04
C02
362
1980
1800
1520
425
2540
2420
1650
- 829
- 2640
- 2500
- 1930
- 525
- 3020
- 3040
- 2710
CO
0.8
1.2
1.5
3.6
0.7
4.7
5.8
7.3
- 1.8
- 1.4
- 2.5
- 4.7
- 1.5
- 6.6
- 8.4
- 12.6
*<
      Failed QA Test
ker  ^eroperatures inside  the house  during the  operation of  the  unvented
the 8eile "eaters (Table  2)  were higher than would have been preferred,  but
W0utside  temperature  ranged  from 1 to  23°C during testing  of  the  C/R
fto er and  -l to  21°C  for the  R/R  heater.   Inside  temperatures varied
    22  to  33°C for  the C/R  heater  and  23 to  37°C for the  R/R heater.

                          TABLE 2.  TEMPERATURES

                                   °C
BEATER
•-->^^
C/R


R/R


RUN
NO.
1
2
3
1
2
3
DEN
28
25
28
28
29
32
- 31
- 33
- 30
- 33
- 33
- 37
BEDROOM
24
22
25
23
24
29
- 26
- 29
- 27
- 28
- 28
- 32
OUTSIDE
1
4
9
-1
2
9
- 20
- 23
- 23
- 16
- 18
- 21
                                  717
-------
                                                                    i
     The R/R unvented kerosene  heater  partlculate levels were 8 ~  *     is
the background levels seen in the house (Table 3); whereas,  the C/R  nea
partlculate levels were about twice the background concentrations.
                     TABLE 3. PARTICIPATE RESULTS
RUN TYPE
Background


R/R Heater


C/R Heater


LOCATION
Outside
Den
Bedroom
Outside
Den
Bedroom
Outside
Den
Bedroom
NO.
RUNS
2
2
2
6
6
6
6
6
6
AVG
(ug/m3)
1,2
28.2
34.1
28.7
278.9
242.4
26.2
68.5
35.7
SD*
1.2
19.5
4.4
7.1
39.6
49.4
19.1
17.1
19.3
&
(<
0
8.7
29.8
17.7
219.2
174.9
0
44.4
12.2
                                                                RANGE
                                                                (Ug/mj)
                                                                      2.*
                                                                     47.7
                                                                     38.5

                                                                     36.7
                                                                    353.J
                                                                    312-7


                                                                     89*. 3
                                                                     62.3
      *SD = Standard deviation
     The organic  compounds  collected  by both  Tenax and  XAD
peared to result  from unburned kerosene. No  other major organic
were identified. The  largest  11 peaks  (heptane, toluene, octane,
nonane, n-propylbenzene, n-decane, n-undecane,  n-dodecane, n-tridecan »
n-tetradecane) collected on Tenax were quantified, and the totals are
in Table 4. As with the  particulate  data, the R/R heater organic
were about 10 times the amount detected with  the C/R heater.

                   TABLE 4. TOTAL ORGANICS ON TENAX
                                (ug/m3)
       HEATER
LOCATION
AVG
SD*
RANGE
C/R

R/R

Den
Bedroom
Den
Bedroom
66
49
820
828
18
13
365
117
38 -
27 -
366 -
679 -
93
66
1259
98/
— -
     * SD = Standard deviation

     The R/R  kerosene  heater had a higher  burn rate  of  fuel  then
C/R heater. The  R/R heater  averaged 3,45  g/min;  whereas,  the  C/R
averaged 2.23  g/min.  The higher  burn rate  of  the R/R heater  may
some of the higher emissions from the R/R heater.
                                   718
-------
                            CONCLUSIONS

pa  *"= unvented kerosene  heaters  did produce emission  concentrations  of
  ^culate higher than found in normal indoor air.  The R/R heater emissions
   worse than those of the C/R heater.   However,  these results are based on
   one heater of each type,  and further investigations should be conducted
  te final conclusions  may be drawn.

                            REFERENCES

   *• Jackson,  M.D.,  Clayton,  R.K., Stephenson, E.E.,  Guyton, W.T.,  and
      Bunch,  J.E.,    EPA's Indoor  Air  Quality  Test House,  1.  Baseline
      Studies.   Proc.  1987  EPA/APCA  Symposium  on  Measurement of  Toxic
      and Related Air Pollutants,  pp. 104-108.

   2* White,  J.B.,   Leaderer,   B.P.,   Boone,  P.M.,  Hammond,   S.K.,  and
      Mumford,  J.L.,  Chamber  Studies   Characterizing  Organic  Emissions
      from Kerosene   Space Heaters,  Proc.   1987  EPA/APCA  Symposium  on
      Measurement  of   Toxic  and  Related  Air   Pollutants,   pp.   98-103.
                                  719
-------
DESIGN OF A SELF-ADMINISTERED PERSONAL
DAILY ACTIVITY QUESTIONNAIRE FOR EVALUATING
EXPOSURE TO COMBUSTION PRODUCTS
N.C.G. Freeman
Graduate Program in Public Health
Dept. of Environmental and Community Medicine
UMDNJ-Robert Wood Johnson Medical School
675 Hoes Lane, Piscataway, New Jersey
J.M, Waldman and P.J. Lioy
Dept. of Environmental and Community Medicine
UMDNJ-Robert Wood Johnson Medical School
675 Hoes Lane, Piscataway, New Jersey

                                                                 • h
    As part of the Total Human Environmental Exposure Study (THEES) whicn
at benzo{a)pyrene exposures in a community directly impacted by a foundry. a
self-administered personal daily activity questionnaire was developed. This
questionnaire was designed to keep an accounting of the microenvironmentai
exposures of the study participants, and components of the form include: (a) °
schedule of activity; (b) home and work place heating and ventilation; (c)
and ETS; and (d) cooking activities.
                                                                rtic
    The questionnaire was designed to be completed on a daily basis by Pan
who, after gaining familiarity with it, can fill out the form within 5 minutes. Tn0  -,05
two-week phases of the THEES study included two adult participants in most n
and questions were included to provide internal checks for response validity-

    The responses were encoded to provide a graphical display of individual
exposure opportunities for conbustion products. This display allows rapid
identification of subjects likely to have high versus low personal exposures on
basis.
                                 720
-------
 da •    ^ota' Human Environmental Exposure Study known as THEES was
 , Sl9ned to look at benzo(a)pyrene exposures in a New Jersey community which
  s  foundry.  During the first phase of the study indoor and outdoor PM-10 and BaP
 k~'s were monitored daily for two weeks and a daily questionnaire covering
   S6hold activities and food habits was completed for each household 1.

 Out Curing phase 2 of the study, upon which this report is based, the household and
  Qoor monitoring were replicated and in addition members of each household
    Personal air monitors for the two week study period "\ A new questionnaire was
 OJQ "?Ped which both replicated the phase one questions and provided details of
 ^ ^icroenvironment of the study participants which aid in understanding sources of
 thftS°na' exPosure to air pollutants^. From the information on the questionnaires
 ^Participants' exposure to two outcome variables, PM-10 and BaP, can be
 atlcjrrecl- The resulting information can be used to produce a form of life-style analysis
   a semi-quantitative index of exposure.

   he
  i ^"he THEES study 9r°up who completed the questionnaire was made up of 13
      als ranging in age from 27-74 years from 8 households. In addition to filling
      questionnaire each participant was required to carry a personal air sampler
     ovide daily urine and food samples for 14 days. Although the study group was
    it reflected the variability to be expected in a community including housewifes,
      and those emP'oyed full-time and part-time. None of the participants were
     rs and smoking seldom occurred in any household.  Several households had
   r)ted space kerosene space heaters in the living rooms and one household
     ed a living room coal burning stove. Three of the homes contained electric
      and the remaining 5 had gas ranges.

   lhe questionnaire was designed as a 24-hr recall diary which was filled out on
    ternoon of each day during the study and collected by THEES field members at
   "[tie. The 7 page long questionnaire, containing 94 response items, was
    zed by topic area so that a participant could easily complete the questionnaire
    5 minutes.  If a participant neither smoked nor was exposed to smoke, the
    devoted to tobacco use and exposure to smoke could be passed by. Design
      of tne questionnaire include brevity, ease of completion by the participant,
       of encoding. Imbeded into the questionnaire were questions which
       internal checks for response validity, while at the same time provided more
%e,  .  information about specific exposure variables. An example of this is the
Hi0 st'°n on duration of travel time in the environmental compartment section and  the
  e ^tailed exposure questions which relate to the types of transportation used
                                 721
-------
and their duration.

    The questionnaire was field tested prior to its use in THEES and modification
were made to enhance question clarity and ease of use by participants. All the
questions were reviewed with the participants prior to the start of the study (at  grfl
same time that the household monitors were being set up and the participants w
being fitted for the personal monitors). This was done to reduce any ambiguity3
how the questions were to be handled and provide training for the participants o
ho,w to fill out or circle their responses.
    The questionnaire was divided into 5 environmental compartments:- horn0
indoors elsewhere, outdoors, and travel.  Each participant marked on a time I"1
amount of time spent in each environmental compartment. Based on these tirjlent$
lines, lifestyle characteristics can be displayed graphically and those compart'y
which were most likely to contribute to the participant's exposure can be pinp°'
    Within each of the compartments further questions focused on relevant <—   0
and factors which might influence exposure. These included smoking and e*p°'  ^0
to smoke, cooking activities, use of appliances and machinery,  exposure to 9a  0
engines, modes of and duration of travel, ventilation and heating sources, and   ^
spent in strenuous activity. Potential sources of exposure were  identified by
regression of personal PM-10 values on to the exposure variables. Those
with regression coefficients >.115 were then used in a stepwise regression
contributions to total exposure (table 1).

    The responses to the questions were organized for numerical encoding f°r ^
analysis by SAS.  For this presentation preliminary analysis of potential sour!c?ca|
PM-10 exposure was done on a Macintosh Plus using the Statview 512 statist'
program.
Results

    A variety of lifestyles was identified among the participants in this study l"^f #
At one extreme was an individual of very regular habits who spent an avera^? s
hours per day at home in sedentary and solitary activities. More typical 'if05^ 0|i/i(J
involved regular work hours during the week with a related regular level of tra  ^
and a variety of activities undertaken on weekends. While these individuals uSk0Sit
wider number of environmental compartments the regularity of their habits ma
relatively easy to identify potential sources of exposure. At the other end of tn
lifestyle continuum were individuals with complex work schedules, irregular tra
habits, and extensive range of activities and potential exposures.

    Unique influences on exposure could often be identified from questionnai
                                   722
-------
 hi^P°nses. For example one participant's PM-10 levels were averaging 127 |ig/m3
  ' * T one day in which they rose to over 900 jig/m3. Analysis identified equipment
    specifically use of an arc welder for 3 hours on the day in question.

   Using stepwise multiple regression, 17 to 93% of personal sampler PM-10 level
    •rally mediated exposure sources could be identified. Since participants in the
    spent on average between 13 and 23 hours each day in their homes many of
   sources of exposure were in the house. Specific activities within the house were
 ! -n better predictors of PM-10 levels than the amount of time spent in the house.
 J|ls may be explained by the lack of variability in the time spent in that location.
 JJ^'fic activities which were related to PM-10 levels included the following cooking
 I lv'ties: time spent frying, roasting, broiling, using of griddles and toasters. Other
 ctors which influenced PM-10 levels were use of unvented space heaters, coal
 h rning stoves, exposure to tobacco smoke in the work place, house cleaning, and
 Urri'ng food.
u  The THEES self-administered personal activity questionnaire has been found to
  a useful tool in identifying sources of personal exposure for PM-10. Further
fj"j%is is being done to assess the contribution of these variables to elevation of
      levels above background rates and to develop a model of personal exposure.
      s in using this instrument depended  not only on the design of the
    ionnaire, but also on the daily retrieval of the questionnaire. Retrieval rate in
""s study was greater than 95%.
^  Based on pilot data the questionnaire is expected to be equally useful in
. a|yzing BaP exposure. While the questionnaire was developed to deal with a
  a" sample study, its design and ease of use makes it applicable to larger studies.
«  Correlations of internal validity questions yielded significant coefficient values for
*n? °f the study Partic'Pants- Further examination of the problems of test validation
  a Participant performance are to be done.

         ements
   The authors thank the other members of the THEES team, in particular Timothy
       Ramana Dhara and Thomas Wainman for their aid and useful discussions.
  are most indepted to the THEES participants who allowed this intrusion into their
  " faithfully wearing their personal air samplers and regularly filling out the
  . 'onnaires. Work reported here was funded by New Jersey Department of
 IVtronmental Protection.
                                 723
-------
References

1.  Lioy, P.J., Waidman, J.M., Greenberg, A., Markov, R., and Pietarinen, C.
"Total Human Environmental Exposure Study (THEES) to Benzo(a)pyrene:     ^
comparison of inhalation and food pathways." Arch.Environ. Health.Cm press, 19

2.  Buckley, T., Waidman, J.M., and Lioy, P.J." High-flow, 24 hour personal
sampling: problems and sQlutions."Proc. 81 st Annual Meeting APCA. (in  press,
1988).

3.  Quackenboss, J.J.. Spongier, J.D., Kanarek, M.S., Letz, R., and Duffy, C.P.  ^
"Personal exposure to nitrogen dioxide: relationship to indoor/outdoor air quality
activity patterns." Environ. Sci. Tech. 20:775-783. (1986).

4.  Wallace, LA., Pellizzari, E.D., Hartwell, T.D., Sparacino, C., Whitmore, R-   xic
Sheldon, L, Zelon, H., and Perritt, R. "The TEAM Study: personal exposure toi
substances in air, drinking water, and breath of 400 residents of New Jersey, N
Carolina, and North Dakota." Environ.Res. 43:290-307. (1987).
                                   724
-------
                            Table 1

-Activities and Environmental Conditions Associated with PM-10

      PIP              Activity/Condition

      01                ventilation, cooking, soldering

      02                ventilation, bus riding

      11                vacuuming, food burning

      31                household cleaning agents

      41                ventilation, housecleaning

      62                household cleaning agents, carpentry

      81                griddle use, furnance use

      82                griddle use, frying

     91                unvented space heater use

     92                welding, petroleum products,
                       occupational ventilation
                              725
-------
   PID62
   100.


1   80

S.
j=   60


*   40.


    20
            OTRAVEL
            OVORK
             DOUTDOOR
             + HOME
    A INDOOR
                              6       8
                                 DAY
                            10
        12
14
  PD92


   100-


    80


    60


    40


    20 J
            OTRAVEL
            OVORK
-------
4i
  i,p\v Cost Data Acquisition System for
  ldential Combustion Spillage Monitoring
  lT-tr
    Uwton' P-Eng-
    n/ Lawton, Parent Ltd.
   Canotek
     , Ontario
^0   In response to a Request for Proposals from Canada Mortgage and Housing
&H h°raticm for research in combustion product spillage from residential appliances,
Hljfr^ Lawton, Parent Ltd. developed a data acquisition system based on a standard
   °"Computer and games card.

in*   The system allowed the continuous monitoring on eight channels of status
sixtritlau'on and the control of two air sampling circuits. Monitoring was carried out on
}^ee^ houses in the Ottawa and Winnipeg areas during the period of January through
^q   1987.  The monitoring system and methodology allowed the determination of the
^   6ncy and duration of spillage, related it to house operation factors and characterized
       on indoor air quality.
                                   727
-------
BACKGROUND

     In a previous research project, Canada Mortgage and Housing Corporation ^%g
surveyed over 900 houses to determine the prevalence of combustion product spilla&
from residential furnaces and hot water heaters. Spillage was detected by a series o
permanent colour change temperature indicators mounted on a card immediately   ^
outside the dilution air inlet of the appliance. Incidences of spillage were recorded 1
significant percentage of these houses, prompting CMHC to undertake more detail
research.

     In November, 1986 CMHC issued a Request for Proposals to undertake
in up to 20 houses to determine  the frequency and duration of spillage occurrences/
these related to building envelope and operating characteristics, and  the effect sp"
had on indoor air quality.  Two major constraints to the project were budget and
schedule.  The work had to be completed in three months from January to March
for Can$75,000.

     Buchan, Lawton, Parent Ltd. was awarded the contract, largely because of t°e
innovative method proposed to  obtain the required information at relatively
The development and application of this methodology is the prime focus of
The results of the study and previous CMHC funded studies on combustion
be obtained from the Research Reports listed in the references (1) (2).
SYSTEM CONCEPT PLANNING

     Given the limited time available to develop a methodology, it was quickly
that the use of "conventional" data acquisition methods was not possible.  It wa
to be too expensive to deal with a reasonable number of houses and too long a *e
was required to get the work carried out before the heating season ended.  Conseq
Buchan, Lawton, Parent Ltd. went back to basics.
                                                                          e
     The first step was to identify the minimum information requirements. Tn
identified as:
  - Real-time, continuous monitoring to determine the duration of spillage and
    related to household operational factors.
  - Household operational factors. Information  required included:  were aU**
                              .
   operating, were they spilling, were any of the doors or windows open, were a /
   exhaust fans operating, and was the fireplace (if any) being used. Information
   however, could be limited to "yes/no" status format rather than analog ir"°     ,j
                                                                        rriflS
  - Measurement or collection of a sample of indoor air when spillage was occur*1*
   some kind of base line to compare to, such as an "average" sample.            /
                                                                        */*/! 0
  - Monitoring systems would have to remain  in the house for a significant p^ .^jjiU^'
   time (since spillage  is weather dependent).  Three weeks  was felt to be the m
  - Limited interior wiring because of cost and aesthetic penalties (private houses   y
   being used and the systems would have to  be moved at least once during the y
                                                                       id ha
  - Ability to use low cost rental equipment. Any purchased components would
   be immediately available.

     Characterization testing of these houses was also required.  Specifically/    ^j
determining their envelope characteristics using fan depressurization methods

                                      728
-------
  ermining levels of contaminants that were generated in periods of forced combustion
^uct spillage.

te ., The identification of these broad principles was one thing, translating them into
kj lty was another.  Credit should be given to Don Marshall of SRO Engineering for his
. . ^ in developing unusual micro-computer applications. The end result was a system
  n8 a standard IBM-PC "clone" micro-computer system and a standard games card.

   .  e key to the data acquisition system was the eight inputs available through the
   tick controller" on the games card.  This allowed four status and four analog inputs.
ort     could also provide the control outputs that were needed. While there were
  er
      tions, the "speaker" output and the request to send (RTS) pin on the serial
   ace of the system were used. A prime attraction of this building block was that any
       number of systems could be rented at very reasonable rates.
^  r op
  rac
       DETAILS

CQ  Once the parameters of the monitoring system were defined, development was
r  ^ntrated On how tO Sense the SDecifiC •n''xrTn^''^rtn rarmiraH ar»H rnnvort if fn i
     that could be dealt with by the PC.
   por gas-fired appliances, it was decided that the best indicator of spillage was a
  Perature sensor mounted immediately outside the dilution air inlet.  High
 ^erature thermistors were used because they were readily available and provided an
   "" that was easily read on the analog channels of the joystick controller (0-100K Q).
     *log signal was converted to status format in software using a field adjustable
   lh°ld value.

inj. A different approach was used for oil-fired units. Previous CMHC experience
doi cated that temperature was a less reliable indicator of spillage. Standard smoke
      heads mounted above the draft damper were used instead.  A simple power and
    conditioning circuit was required to get a reliable output that could be read on a
    channel.
Sic
     he most obvious method of sensing the on/off status of the appliance would have
   a relay in the thermostat of the unit.  That would, however, present some
^lications. In many fuel-fired hot water heaters, the thermostatic control is built in
'ify*10* easily accessible.  Secondly, there were concerns with liability issues. If possible,
S? desirable to keep the monitoring system non-intrusive. It was decided to use a
  Pfirature sensor mounted in the flue upstream of the dilution air inlet.

   Initially, concerns were expressed about whether thermistors could survive
   h  ted flue 8as temperatures and contamination. For that reason, the search for
   er sensor was initiated. Some experimentation indicated that a simple and
S|a Pensive solution was immediately available.  It was found that the resistance of a
     d 100K resistor varied enough over the expected temperature ranges (90%) to
       determination of operating status. This "sensor" was used for the project but it
C *<* less than fully reliable.  Further experimentation has shown that the high
  Pfirature thermistors used for spillage indicators can, in fact, survive flue conditions.

     itn only eight input channels, it was not possible to instrument each door and
   ,°w- This, however, was not critical. The real concern was whether the house was
   sealed" condition and, secondarily, if there was an intentional opening,  whether it
                                     729
-------
was on the windward or leeward side of the building. Non-contact magnetic
were mounted on the exterior of operable windows and doors. All wiring was run
outside the building and brought inside to the PC through a basement window. Tn   j
switches were hooked up in parallel on two circuits, one for the north and west si<*e
one for the east and south sides of the house.

     Fireplace or wood stove operation was sensed with a thermostatic switch nio
four feet down from the top of the chimney on a length of pyrotenax cable. Again/
wiring was external.

     Dealing with exhaust fan status was a little more difficult. There can be a nu01
of them in the house and any direct connection would require internal sensor win B"
The local Radio Shack supplied an answer in the form of wireless intercoms. At &
fan, whether it was a dryer, kitchen fan or bathroom fan, an intercom with its "scn.atej
switch strapped down, was hooked up in parallel with the fan power circuit. Any  . ^
fan  was turned on, the intercom would send a signal. A receiving unit was moun
the box with the PC. The speaker output of the intercom, with some conditioning/
input to a status channel.  It was also found necessary to put in a software filtering
element but, once this was done, the system worked remarkably well.
AIR SAMPLING METHODS

     The air sampling aspects of the project were intended to characterize, rather    j
scientifically quantify, the impact of combustion product- spillage on indoor air qua J
relied primarily on bag samples of air from a location near the furnace.  It was ass
that, if high levels of contaminants were detected at this "worst case" location, m<>
extensive sampling could be carried out but, if not, extensive sampling would be
wasted effort.

                                                                       n**^
     The air sampling packages used standard, diaphragm-type aquarium pump3
were activated by the PC via solid state relays and collected samples in 10 litre,    .^ ;
aluminized mylar bags. Two control circuits were used. One initiated sampling l
field adjustable delay) when spillage was detected and stopped the sampling &&**.
another adjustable delay) when spillage ceased. The other circuit was a timed eye*
which commenced after the first spillage incidence and activated the pumps for 3
seconds each half hour.

     NO/NO2 sampling using Draeger sampling tubes was also carried out in so
houses.  This used a separate set of pumps on the same circuits.
     For the bag sampling approach, the sampling flow rate was not critical excep
means of controlling the sample size.  For the rube-type sampling, the sample size ^
critical.  In either case, the flow was field measured and set with a rotameter and
valve. The sample size could be calculated from this flow and the sampling
recorded by the PC with an error of ±20%.
SOFTWARE AND DATA STORAGE

     The data acquisition software was written in BASIC by Don Marshall. It
sequentially scanned the eight input channels and the dock. Analog signals
compared to the preset thresholds for conversion to status output. Data was ony
to disk when a change in status was noted.  At which time the status of a

                                      730
-------
.   «me was recorded.  The program also controlled the air sampling package.  An
  walizing routine allowed field setting of on delays/ off delays and sampling periods as
 eU as the threshold values used for analog to status conversion. The current  values of
 * status and a running total of the pump run time was displayed. The system would
       if there was an interruption, losing a minimum of data.
„   --«« data from the system consisted of lines of data showing the time and 1 or 0 for
 e status of each channel. Since this was somewhat difficult to interpret visually,
 °tner simple program was used to scan the entire data set, storing only the data
         with spillage incidences.  This made totalling the number and duration of
      incidences relatively simple.  The effects of aggravating factors, such as fan
      n and mitigating features such as open windows, were determined statistically
  another program.
^o METH
          ODOLOGY
    For the CMHC project, a total of sixteen houses were monitored. The original plan
   to monitor twenty in the Ottawa area but, in order to find accessible houses with
   Wn spillage problems, it proved necessary to extend the search to Winnipeg. In the
 a' nine houses in Ottawa and seven in Winnipeg were monitored.

^   During the installation visit to each house, a data acquisition system was installed
det! c°mmissioned and the characterization testing carried out.  This included a fan
.Pfessurization test and some air sampling under forced spillage conditions using tube-
 Pe samplers . The work was done by two field technologists in one day.

    Over the following few weeks, the house was visited weekly to collect the data
    and ensure that the system was working properly. Air sampling packages were
^ y installed in houses where naturally occurring spillage was detected.  While these
If j  * to place, the homeowner was phoned each day to see if any sampling had occurred.
^lih    a tecnnologist .went out, collected the sample and changed the disk on the PC.
 ^-averaged tracer gas testing was performed concurrently with the air sampling.

CQ   The bag samples and the NO/NO2 tubes were sent to a lab for analysis. The
^Centrations of CO, C02/ methane and non-methane hydrocarbons  were determined by
fjChrornatography from the bag samples. For a limited number of samples, VOC
            by mass spectrography was also done.
     EXPERIENCE
    ,  ,      .   ,
    Usable data was produced for 322.
     s alluded to earlier, many of the problems encountered were with the resistors
    '0 determine appliance on/off status. With the benefit of hindsight, high
     ature thermistors would be recommended if non-intrusive sensors were
      .  ft Would also be possible to use a relay at the thermostat.
           problem was related to the air sampling rates and times. Obviously a
   Ce had to be acheived between collecting enough of a sample to be usable but not to

                                   731
-------
exceed the capacity of the sample bags.  Of course, spillage doesn't work to a set scheo
and occasionally incorrect estimates were made.
                                                                             -I.«
     The third problem worth noting was primarily a developmental one related W
inadequate filtering or conditioning of signals.  Occasionally, the status of a channe
flickered on and off.  This could fill up a disk in less than a week. The problem
limited by incorporating additional software filtering elements.
SUMMARY OF FINDINGS

     The following is a summary of the findings of the project. For the complete f
the report can be obtained from CMHC. (1)
     Of the sixteen monitored houses nine showed no spillage activity during
period of monitoring. Five gas heated houses had spillage incidences of over ten
seconds.  The remaining two houses, which were oil heated, had brief,  infrequent
periods where spillage was detected.
                                                                           r^fl
     In two of the houses which showed significant spillage occurrences, spillage
found to correlate with the operation of exhausting appliances - in one case a fireP (
and the other an exhaust fan.  In the other three, spillage occurred regularly with"
these aggravating factors, indicating poor chimney operation.

     No direct correlation could be seen between spillage frequency or duration an
weather data obtained from the Atmospheric Environment Service.  This is not    p
that there was no correlation but, rather, that other factors, such as appliance ope
or chimney action, overwhelmed the weather effects.

     The air quality sampling showed some increase in contaminant levels attri01
to combustion spillage but the increase was not dramatic. During forced spillage
some contaminant levels, particularly carbon dioxide, were well above ambient 1
                        to 6600 ppm). However, over the loner term monito**
(C02 concentrations of up to 6600 ppm). However, over the longer
even in those houses which spilled consistently, the contaminant levels of satnp
taken during spillage were  well below Health and Welfare Canada's guideline Ie
long term exposure. (3)

                                                                        .
     Taking into consideration that the sample selection was limited to houses
suspected of spillage activity, few had significant spillage incidences. It was not
well, that the frequency and duration of spillage and the levels of contamination
very house specific. While hazardous levels of contamination were not
naturally occurring incidences, there is reason to suspect, in a limited number o
that contamination levels could be a problem. Results are, however, comforting
they indicate that this  should be rare.
REFERENCES

(1) CMHC. Residential Combustion Spillage Monitoring, by Buchan, Lawton, Parent Ltd- *
     Engineering Ltd.  Ottawa, Ontario: Queens Printer, 1987                          «

(2) CMHC. Residential Combustion Venting Failure - A Systems Approach, by Scanad* $
     Consortium. Ottawa, Ontario: Queens Printer, 1987

(3) Health and Welfare Canada. Exposure Guidelines for Residential Indoor Air
     Ottawa, Ontario:  Queens Printer, 1987.

                                      732
-------
  r P0und  and  Status of Computerized Systems
         Title  III Emergency Planning
    Accidental  Chemical  Releases
 j C- Bare
 Mr" Envlronmental  Protection  Agency
 Steg and Energy Engineering Research Laboratory
   arch Triangle Park, North  Carolina  27711
[T  The Emergency Planning and Community Right-to-Know Act of  1986
te»  e Hi of the Superfund Amendment and Reauthorlzation Act (SARA)]
H$g  res facilities handling any of the designated chemicals  [Extremely
^an   Us Substances (EHSs)] in quantities greater than the Threshold
1es   nS Quantities (TPQs) to submit information to their State Emergency
(ljEpptl'3e Commissions (SERCs).  Local Emergency Planning Committees
ftojj, !   engage these facilities in planning and request information
W    e facilities that is necessary for planning.  LEPCs are respom
b   Wi* 't H 4
y Q  lclng emergency response plans for dealing with chemical accidents
^cj tober 1988,  and reviewing these plans annually.   A user-friendly
^Hd]  °S^ comPuterized system has been developed which allows LEPCs to
Pote ! tae large  quantities of data and assists them in analyzing the
       ^  nazard  of each chemical by assessing the severity of the
     Uences  °^ a Pre~plannetrate  on hazards analyses for emergency planning for  accidental
        of EHSs.
                                 733
-------
INTRODUCTION

     With the passage of Title III of the  Superfund Amendment and
Reauthorization Act on October 17, 1986, many  changes occurred in
Community Right-to-Know and Emergency Planning.   Section  301 required
the governor of each state to appoint a  State  Emergency Response
Commission  (SERC) by April 17, 1987.  The  SERC is  responsible for
appointing  Local Emergency Planning Districts  (LEPDs) and appointing*
supervising, and coordinating Local Emergency  Planning Committees
(LEPCs). (1)

     Under  Sections 302-303, any facility  which stores, manufactures,
processes,  uses, or otherwise handles at any time  one or  more of the
Extremely Hazardous Substances (EHSs) in quantities greater than their
Threshold Planning Quantities (TPQs) as established by the U.S.
mental Protection Agency (EPA) is required to  report to the SERC.
chemical is identified to the LEPCs during the development of the
(in accordance with the trade secrets provisions of Section 322).
tional data may be requested by the LEPC under Section 303(d) which
"requires facilities to provide the committee  with information
development or implementation of the local emergency response plan.  *

     The National Governors' Association's "Interim Report:  The States
Designation of Local Emergency Planning Districts" published in August
1987 stated that 49 states and 1 territory had met the July 17, 1987,
deadline for forming LEPDs.  Thirty-five states designated political
subdivisions (counties, municipalities, or a combination  of the two)   ,
LEPDs; 10 states named regional planning or response areas as LEPDs;
5 states designated the entire state as an LEPD. (2)

     The LEPC emergency response plan must include:  "identification °
facilities  and extremely hazardous substances  transportation routes;
emergency response procedures, on-site and off-site; designation of a  .
community coordinator and facility coordlnator(s) to implement the pla 'e
emergency notification procedures; methods for determining the occurre
of a release and the probable affected area and population; descript*0.
of community and industry emergency equipment  and facilities, and the
identity of persons responsible for them; evacuation plans; descript*°
and schedules of a training program for emergency response to chemicfl
emergencies; and methods and schedules for exercising emergency resp°n
plans." (3)

     To assist the LEPCs in writing their communities1 emergency resp° ^
plans, the National Response Team (NRT) published a document, "Hazard0
Materials Emergency Planning Guide" (NRT-1), which was required under
Section 303(f).  This document Is available free of charge from:   the
Emergency Planning and Community Right-to-Know Hotline (800-535-0202)- j,
The "Hazardous Materials Emergency Planning Guide" describes "how to *
a local planning team,  find a team leader, identify and analyze hazar
identify existing response equipment and personnel, write a plan, an«
keep a plan up to date."  (3)

     To assist LEPCs in focusing on the chemicals and facilities of c
most immediate concern from a community emergency planning and resp°n
perspective, EPA published the Extremely Hazardous Substances List an
                                   734
-------
j  eshold Planning Quantities.  This list originally appeared in the
 e
-------
of each screen, allowing the user to move to any database or operation*
The IBM compatible version is still in the development stage. Featu?63
will be listed in "menu format," allowing the user to move to appHca
databases or operations.

     Each step in hazards analysis can be addressed using the systems
presented.  The three basic steps are:  hazard identification,
vulnerability analysis, and risk analysis.  Each step builds on the
previous step to understand the hazards.  The LEPC can analyze the
of the 362 EHSs or of a much larger collection of chemicals.  The
computerized systems are designed to address only the EHSs, but may be .g
modified to include other chemicals.  The three steps to hazards anal?3
are discussed below.
HAZARD IDENTIFICATION

     Hazard identification involves identifying information about!
chemical identities, the location of the facilities, the quantity of
material that could be involved in an airborne release, and the type °
hazard.  Under SARA Title III, facilities handling any of the 362
were required to report the chemicals they were handling and their
inventory quantities.  The chemical, physical, and health effects      .
properties of these chemicals are already documented in the EPA Chemica
Profiles [issued in November 1985 as part of the Chemical Emergency
Preparedness Program (CEPP) Interim Guidance (5)].  It is the          y
responsibility of the LEPCs to determine what additional information £
would like to have and request the facility to submit it to them.  ^  ,ng
may request any information they deem necessary to complete their plan
process under Section 303(d).

     Both computer systems assist in hazard identification by
a storage location for facility and chemical data reported by
under Section 302.  Separate databases within the system also allow
storage of chemicals reported under Sections 311-312.  This informat*0
may be edited, deleted, sorted, and searched to provide reports of
specific chemicals in a locality, specific facilities in a locality* °
various other combinations.
VULNERABILITY ANALYSIS

                                                                      V
     "Vulnerability analysis identifies areas in the community that iaa'
affected or exposed, individuals in the community who may be subject
injury or death from certain specific hazardous materials, and what   ^
facilities, property, or environment may be susceptible to damage sho
hazardous materials release occur."  (4)
     Vulnerability analysis takes information from the hazard
tion and analyzes it.  Having the chemical, physical, and health effeC
data of a chemical available on the computerized systems, the maximunl
quantity of the chemical which could be involved in an accident, and   Q
equations to determine the dispersion of the chemical, it is possible
determine the concentrations of the chemical at various points around
facility.  Knowing the concentrations and the health effects data, *c  -
possible to determine the vulnerability zone.  After Identifying the a
that is likely to be affected, the vulnerable population of a screenin»


                                   736
-------
     release  is  determined.

     Simplified  dispersion models  and  default  values  for source strengths
  * meteorology  are used  to  calculate  the  vulnerable  zone.  These
   Plifications  speed up  the process of  calculation and reduce the amount
   information and technical calculations  required.  They also place all
   the scenarios  In the community  on a common  base.  The answers received
  Ola these calculations are  approximations and should be used for relative
  rPoaes, not for real-time  decisions  of evacuation or shelter-in-place.

 ^    Dispersion  models may be familiar to  some members of the LEPC.   These
   els are often  used in  routine releases  or emergency response computer
 traPhics.  The familiar shape of the plumes resulting from a release Is a
 *ardrop.  For purposes of planning, this  plume is  rotated  360 degrees,
 p.11^ It Is Impossible to determine the  prevailing  wind direction when
  anrn«~ for  an unexpected release.  This  entire circle is  considered the
          zone.
j   These circles should be  plotted on  a  scaled  map  of  the  community to
  8Play the areas that could  potentially be  affected.  These maps  will
.  Lst the community In their planning  efforts  and  in  completing  the  risk
  **ysis — the final step in hazards analysis.
    ANALYSIS

c   Risk analysis Is composed of two basic  factors:   probability  and
uj*8equences.  "Risk analysis is an assessment by  the  community  of the
  *elihood (probability) of an accidental release  of a hazardous material
^? the actual consequences that might occur, based on the  estimated
 U1*erable zones."   (4)

i   The vulnerable zone should be analyzed  to determine the impact of
8,Irig encompassed by an acutely toxic cloud.  The  major factors which
t °uld be considered and which affect consequences are:  resident  popula-
  te within the zone, transient populations, sensitive facilities/
p0Pulations, and other special conditions.   Transient  populations  are
QJ>uiatlons who mav or mav not live w;tthin the zone, but who may
      the zone at a particular time.  These would include  residents of:
     rs, subways, shopping centers, airports, offices, churches,  sports
     ',  public parks, and other public facilities.  Sensitive populations
a*e Populations who are more vulnerable or more easily affected or who
8ch Cr*tlcal to maintain intact.  These would Include  occupants of:
8(. °°l8, hospitals, nursing homes, day-care  facilities, prisons, police
Unions, and fire stations.  Other special  conditions are  those which
h  affect consequences.  An example of a special  condition would  be the
  r evacuation routes accessible to residents of  an Island.

-------
probability.  These rankings may be a simple High, Medium, or Low ranW
for both variables.  When the ranking is completed, the community will
able to address more appropriately the process of planning for these
releases.

     Risk analysis, which will Involve opinions of members of the LEPC ^
determining the final priority of each facility, is a less definitive a
more subjective procedure than the first two steps.  It involves much
local individual research, which may include a review of the facility   f
and/or the facility's hazard evaluations.  Both computer systems allo*
storage of the final priorities for consequences and probability and
discussion of how these priorities were determined.

     The plans need to be updated annually.  Both computer systems si
for re-evaluation of the vulnerable zones and reviews of the priorlti
set in risk analysis.
SUMMARY

     User-friendly computerized systems are being developed by the EPA  g
in collaboration with the NOAA, which will allow LEPCs to handle the 1*
quantities of data expected under SARA Title III and assist them in
analyzing the potential hazard of each EHS by assessing the severity ° .j
the consequences of a release for emergency planning.  These systems   .
assist In the three steps which are involved in hazard analysis:  naz
identification, vulnerability analysis, and risk analysis. These syste
initially use many assumptions in dispersion modeling to determine tn ^
preliminary zone of potential impact.  These areas may be re-evaluate  .8r
refined later using fewer assumptions.  The Macintosh version is very
completion, and the IBM compatible version is still under development-
REFERENCES

1.  Federal Register, "Extremely Hazardous Substances List and Threfln°
Planning Quantities; Emergency Planning and Release Notification
Requirements; Final Rule," Vol. 50, No. 77, p. 13378, April 22, 198/.

2.  National Governors' Association, "Interim Report:  The States'
Designation of Local Emergency Planning Districts," August 1987.

3.  National Response Team, "Hazardous Materials Emergency Planning
Guide," March 1987.
4.  U.S. Environmental Protection Agency, the Federal Emergency
Agency, and the U.S. Department of Transportation, "Technical
for Hazards Analysis, Emergency Planning for Extremely Hazardous Sub
stances," December 1987.

5.  U.S. Environmental Protection Agency, Office of Toxic Substances*
Washington, D.C. , "Interim Guidance:  Chemical Emergency Preparedness
Program," November 1985.
                                   738
-------
       GASEOUS COMPOUNDS IN
   '°NMENTAL TOBACCO SMOKE

 .J
 .4' fatough, K. Wooley, H. Tang,
      ls and L'D- Hansen
        Department
 tov   YounS University
   °- Utah 84602 U.S.A.
    :|"ent of Chemistry
   °rnia Polytechnic State University
   ils Obispo, California 93407 U.S.A.
 ,
^    y
      n°lds Tobacco Company
       and Development
     -Salem, North Carolina 27102 U.S.A.
    /]- v
      onmental tobacco smoke  is  an aerosol consisting of  both vapor and
     *Qte phases  and  many organic compounds associated  with the aerosol
    Und  in both  phases.   The presence of  these  compounds  in both phases
    Cates the  sampling  of environmental  tobacco smoke since  the  gas-
           distribution   can  be   altered   by   the   sampling  procedure.
     tations  of  nicotine,  3-ethenylpyridine  and  other basic nitrogen-
 tido0   ng organic  compounds  have  been determined  in  both chamber  and
31<1   environment  sampling experiments.   The  results show  that  nicotine
\es  r  tobacco  alkaloids are  present in  both  the gas  and particulate
\   °f  environmental  tobacco  smoke and that their  distribution between
J5tti W°  Phases  is  variable  in  indoor environments.    Since the  vapor -
SjHj  ate distribution of these  compounds  may be altered as  a  result of
%js ^B>  the experimental  determination of the phase distribution  in  the
!Stn   8 may be dePendent on the  sampling  technique used.   The results
 %     by sampling environmental tobacco smoke in a  30-m3  Teflon chamber
^ben iffusion   denuders,   passive  personal  monitors,  filter  packs  and
  6rW  beds  have  been  comPared-   The  resulcs  suggest  that  accurate
  Q Ration of  both  the  gas  and  particulate   phase  concentrations  of
     ds   in  environmental  tobacco  smoke  can  only  be  accomplished with
     g  systems  where the gas phase material is collected  in the  presence
     Patticulate phase  compound.
                                  739
-------
 Introduction
                                                                     ,
    The loss or gain  of significant amounts of  semivolatile  compounds   ,
 particles during  sampling causes  errors in  the  determination of  aei: ,of
 chemical  composition.  This problem was first encountered in the colleC  ^
 of gas and particulate phase nitrate because  of the  semivolatile  natut
 NH4N03 .   Extensive  studies have since  shown that reliable  values
 V>ncentration of  particulate  nitrate can  only be obtained by the
 diffusion  denuders   to   collect   HN03(g)   prior  to  collection   o    ^
 particles .    Similar  problems  must  be considered  in  the  collecti°n ^
 semivolatile organic  compounds.    Accurate  collection  procedures  f°r ,jg
 determination of both gas  and particulate phase components  of  semivol3
 organic compounds  must meet the  following two criteria:

    1.   Organic  compounds  initially  present in the  gas  phase  must
    adsorbed onto particles or  a  particle collection filter during  sam
                                                                       ^
    2.   Organic  compounds  initially  present in the particulate phase
    be  captured separate from compounds  which are present in the gas  P*1

                                                                      h ^
    These  two criteria  cannot  be met by a  sampling procedure  in whic  ^
 particulate  phase  is  collected  before  the  collection  of compounds *• -^
 gas  phase because  the gas  phase  compound  and  compounds volatili26**.^
 particles become  indistinguishable.   If organic compounds are volat*   5
 from  particles  during  sampling,  it is  necessary  to  first collect  t*ieaflic
 phase  organic compounds and then to  collect the particulate phase  ofe ^
 compounds with a sampler which will  collect  all  organic  material, be
 or particulate.
                                                                       V ^
    The  unambiguous collection of  gas phase  semivolatile compounds *
accomplished  using diffusion denuder sampling systems.  Several  st
the  development  and use of active diffusion denuders  for  the  coll
gas  phase  organic compounds in  the  atmosphere have been  reported   •   tM
laboratory  development of annular denuders  with activated charcoal a  Jfi
sorptive  surface7 and  of cryogenic  traps8'9  for the  sampling of °(^i<^
vapors has  also  been described.   Sampling systems using sorbent  beds   y
may  collect  gas phase  organic   compounds  in environmental  tobacco   .^
without removing particles have  also  been described1^'11.   Passive s?2 ,/i^
which collect gas phase compounds by diffusion  to a  sorbent  system    At
also  be effective in  the  collection of  semivolatile gas  phase °  Ap
compounds  without sampling  errors due  to   the  volatilization of roa j
from particles.   Such  passive  sampling devices  have  been used  to c
gas phase nicotine in  environmental tobacco  smoke-*1 -^ >^ .

     Environmental  tobacco  smoke is a  major  contributor to
pollution  in  environments where  smokers are present^-^.  The identi^1  ^
and  quantification of  environmental  tobacco  smoke exposure  is inlP  $\r
because  of  irritant   and   suspected  health   effects  associate^   ed>
involuntary  exposure15'16  and  the  large   population  which   is e3t?^(
Tracers of  environmental tobacco smoke used in the past include  fe5^ fl*'
(or  total)  suspended   particulate   matter  (RSP) ,  CO,   nitrogen  ° Q^
nicotine,    3-ethenylpyridine ,   solanesol,    N-nitrosoamines ,   a  ce**'
hydrocarbons, acrolein and frequency  of smoking.  Of these various tf^e l
only  nicotine,   3-ethenylpyridine15^"17 and  solanesol18'19  are  «n  $$ ,
environmental tobacco  smoke.  The alkaloid  bases in environmental t .
                                                                      .
smoke  are  distributed  between  gas  and  particulate  phase  spec ye&°
illustrated  in Figure  1  by  data from  experiments  conducted  in  a $e  t
chamber17.     3-Ethenylpyridine   is   found  only  in   the   gas  P*1* g  *
environmental  tobacco  smoke  and is  present  in  indoor  e

                                   740
-------
              Which are easily measured17.  Solanesol  Is  found only in the
 P^tic,°nS Wc   are  easily measured.   Solanesol Is found only in the
 "*lrnn   e  Phase in  ETS and  ls  easily  determined  in  environments  where
 "tcoti   ntal  tobacco  smoke is  present18-  BYU unpublished  data    while
 tlle us"6 ^  present  ln relatively high concentration  where  smoking occurs,
 Coaipli       nicotine  alone  as  a tracer of  environmental tobacco  smoke  is
      ad  by  the facts  that   X)  nicotine is found in both the  gas  and
            Phases  in  indoor   environments5'17' 2°.21,   and 2)  gas  phase
    i     S  probablv  rei»oved from  the  environment at  a faster rate  than
 He5,l7te  nicotine  °r  the particulate  portion  of  environmental  tobacco
 6!tPoSur  '   Thus' the concentration  of gas phase nicotine may underestimate
 ^oti S  t0  envlronmental  tobacco  smoke15.    In order to  determine  if
        1     aPPr°Priate marker for environmental tobacco  smoke exposure,
    at                                                                    ,
 Pattlcin' sensitive  methods are needed  for  separately determining gas  and
      iate phase nicotine  in indoor environments.

        S  are  reported  ln  this  PaPer  for  the  comparison  of  several
        Cechnlclues  for  determining gas and  particulate  phase nicotine  in
              tobacco   smoke   generated  in  a   30  m3   Teflon  chamber.
     of     °f  the  results for  sampling for  these  species  with  several
      1 sampling systems are discussed.
     r Sampling Experiments.
  Euvi
 Coi&bUst:j0nmental tobacco smoke was  generated in a 30-m3 Teflon  chamber by
 t%22.   Two sequential
       601-10118 were  used to check  for complete collection of gas phase
1i c°atS by the denuder surface.   The  deposition  pattern of nicotine in a
*  fcln    cylindrical denuder has  been previously studied  to verify that
^ItoJ" is  collected by the denuders with  the  expected efficiency23.  The
°^ect jntal  tobacco  smoke was  sampled at 20 slpm.  Duplicate samples were


 Sc
             R<*'ria--L   ^as phase nicotine was  collected using a previously
                 --
               sorbent  sampling system10.  The  collection efficiency of
 N|Kjf^  ent bed  for  gas phase nicotine  has been shown  to  be greater than
  tet   PUDilshed  data    In  addltion>   use  of  a condensation  nucleus
^ Su  ° monitor ETS  aerosol  with and without an XAD-IV  sorbent tube in
j| ' lv ^6sts  that  smoke particles  are  not  efficiently collected  by the
^6t:ev. °rbent bed.   In the experiments conducted in this  study,  a 25 mm
    tti  P1uoropore  (1pm pore  size)  filter  (Millipore  Corp.) was  placed
     6         sorbent tube to sample particles  passing  the sorbent bed.
                                                                        .
K^en,  tonmental  tobacco  smoke  was  sampled  at 2 slpm  with this  XAD-IV
J"  s^ /our replicate  samples  were collected for  each experiment.   Two of
     P  s  were  analyzed at  Brigham Young University  and two  of  the
            analyzed  at  R>Jt  Reynolds.   Gas  phase nicotine  was  also
       W^tn two different XAD-II  sorbent  bed sampling systems with the
      ntal  tobacco  smoke sampled at  20  slpm  for these systems.  The XAD-
                                  741
-------
 II  In  one sampling system was preceded by an  acid-washed  quartz flit®* ^
 followed  by  a BSA coated filter.  In the other  XAD-II  sampling system,
 XAD-II was  followed by a BSA saturated filter but not  preceded by
 filter.   Replicate  samples were  collected for  each XAD-II  sampling sys1-""
                                                                        FpA
    Passive Sampling Device.   Gas phase nicotine was sampling using *-"  ,$
 personal   sampling   device-'-2  by   placing   a  filter   saturated   ^g
 benzenesulfonic   acid  at  the  collection  area  for  the   sampler •    ej
 effective sampling rate  for the passive collection device was  calcu  ^
 based on  the  diffusion coefficient for nicotine" and published data °n
 passive sampler  »^.

 Sampling  for  Farticulate  Phase Nicotine.
                                                                      f v^
    Particulate phase nicotine was  collected using a filter for  each °  fp
 active sampling  systems described above.  In the annular diffusion den   ?
 the  particulate  phase nicotine  was  collected on  an  acid-washed 4  .^
 filter followed  by  a BSA saturated glass fiber  filter  to  trap  any nic°were
 lost  from particles collected by the  quartz  filter.   These filters      ^
samples  analyzed  at RJR,  nicotine  collected  on  the   XAD-IV  rest*1  ,v)
Fluoropore filters was extracted with ethyl acetate containing ^-^\r  j$
triethylamine.    Quantitation  was  performed  by capillary  column
nitrogen-selective  detection  (NPD)  with qulnoline as  internal
Desorptlon efficiency determinations  and other details of the
protocol have been described^.

Results and Discussion

     The concentrations  of nicotine determined  using  the various
techniques are summarized in Table I for gas phase nicotine and In
for particulate  phase nicotine.   The standard  deviation of  an ob£
from  the mean for the duplicate analyses  indicates that the precis
the various techniques was better than 10 percent.
                                                                      *?.
   The analytical results obtained by the two independent laboratori6^
in  agreement  as  illustrated  by the  results given  in  Figure  2
determination  of  gas  phase  nicotine  using the  XAD-IV samplers.    g
regression analysis  of the data  shown In  Figure  2  (r2 -  0.976)  61  ^
intercept  of  0±328  nraol/m3   and  a  slope of   0.9210.02.    The  a^  i*
difference between  the  results  of  the  two  laboratories,  Figure
5.8±6.1%.  One data pair that differed by 38% is not included.
                                                                       jVi
   The concentrations  of gas  phase nicotine determined using  the   ^0$
the two XAD-II, and the denuder sampling systems were in agreement &s $>
in Figure  3.    Linear regressipn analysis  of  the XAD-IV and denude

                                   742
-------
     ^r2  "  °-968)   an  intercept  of  -122+437  nmol/m3  and  a  slope  of
     '^8.   These data suggest  that  the  XAD-IV sampling system may not  be
         Parfciculate phase nicotine.  The sensitivity of this  test is not
   . enough,  however,  since the particulate  phase  nicotine is only  about
     fche  total  present.   The two XAD-II samplers gave equivalent  results,
   ye 3,  and there  is  no indication that putting  the  filter in  front  of
     "** aorbent bed gave higher results due  to  loss of nicotine  from the
   1     On  tne filter •   However,  this conclusion will also be limited  by
    °W  Percent  of particulate  phase  nicotine present.   Linear regression
       of the  data  in  Figure 3 for the XAD-II sampling systems gives (r2
       an intercept of -9±220  and a slope of  0.9210.04.
^rtii  "
   °
«Kp.
    1 results obtained with  the passive sampling device were in agreement
    those  obtained  with  the  active  sampling  systems  only  for  the
    "Bents where  the  amount  of nicotine actually collected by the sampler
 'Ii6  ar8«, i.e.  for  the  two hour-long sampling experiments, Figure 4.  For
 CK  exPeriments  with high nicotine  concentrations  (combustion of four
 ^s  ^ttes) and  sample duration  of  only 15 minutes,  the  concentration of
 Vit ase  "icotine  determined  with the  passive  sampling  device  was only
 tot  20% of the concentration determined with the active sampling systems.
    ^e  experiments where  the  nicotine  concentration  was  low and  the
    ng  time was  short,  the  nicotine  was  easily  determined  but  the
 tg -ed  concentration  was  very   low  compared  to  the  concentrations
 ^^ined  with  the  other samplers.    These  results suggest  that  the
 CoM6ss ateel  passive  sampling device  itself  may remove  some  gas  phase
     2-  This  is consistent with  the observed rapid removal of gas phase
     a in a stainless steel chamber2 .

     concentrations  of particulate  phase nicotine  determined with  the
      active sampling  systems are compared  in Figure 5.   It  is assumed
        results  obtained using  the  diffusion  denuder  sampler will  be
       because the gas phase  nicotine  is  removed from the sampled air in
 »v;0     denuder  sections22  before  the   particles  and  any  volatilized
 ^e rl   are  collected on the  quartz  filter and BSA  saturated  filters  of
 Mue6tluder  filter  pack.   The  denuder sampling  system gave the  highest
 *te0 ® for  particulate  phase nicotine.    The  results  obtained from  the
 ^  llte collected on  the  quartz filter which preceded the XAD-II  sorbent
 W8p e concentrations of  particulate phase nicotine  which were generally
 %. fchan those obtained using  the  annular denuder  system.   This suggests
 VtnS°me  of  the  particulate  phase nicotine may  be lost  from  the  filter
    8 sampling.

 HD.J the sampling system where  there was a BSA saturated  filter after the
 \ti sorbent   bed,  the  determined  concentration of particulate  phase
    ne  was  always  much  lower than  that  determined with  the  annular
         The  BSA  saturated  filter  would have  collected any  nicotine
          from collected  particles.   The  low results obtained with this
    n8  system  indicate  that  a significant fraction of  the particulate
   * is trapped in the XAD-II bed  at a flow  rate of 20 slpm.   The filter
     after the XAD-IV  sorbent bed also  gave  very low concentrations  of
   Culate  phase nicotine.  However, since  this  filter was  not coated with
    *t  is not possible to determine  if  the observed  low  concentration  of
          phase nicotine Is  due to collection of particles in  the XAD-IV
f^til   "~ or to loss  of  Particulate phase  nicotine  from the filter.   This
\  H,   at^on process  would be expected to be enhanced over the  loss seen
    ;«  quartz  filter before an XAD sorbent bed  since the filter placed
        sorbent bed is  in  an  environment with no  gas  phase nicotine.


                                  743
-------
Summary

   The results of this study suggest that particulate phase nicotine c   j
be accurately collected with only a filter because of loss of nicotin6  ^
particles during  sampling.   An acid coated  filter  or filter pack co*1
used  to  collect total  nicotine,  but could  not be  used  to
between gas  and particulate phase  nicotine.   Data  obtained with a
pack sampling  system may not accurately  represent  the gas and part*-0
phase concentrations of environmental  tobacco smoke  in indoor  environing
The results of the study reported here also indicate that collection °
concentrations  of  gas  phase  nicotine  using  a  passive  sampler ^ •*
affected by  adsorption of gas phase nicotine by the sampler.   The *e  j,
are inconclusive  as  to whether or  not particles are collected by ^ fte*
Reynolds  XAD-IV sampling  system-'-".   These  points  are  all being *£* _ at
studied in  a  second  intercomparison  study conducted  in December 1*  ^u
Yale University  with cooperative participation  by  research personnel ^,
Yale University, Brigham Young University, the University of Massachus
Harvard University and R.J. Reynolds.

Acknowledgement

     R.J.  Reynolds  Tobacco, USA and  the Center  for  Indoor  Air
provided  financial  support of   this  research  through  a  grant to   tf,
Scientific  Inc.    Appreciation is  expressed  to Fern  M.  Caka,
                                                                   J°
               .                                           .
Crawford,  Laura  Lewis,  Galen  Richards,  and  Katherine  C.  Mai°10
technical assistance.

References

 1.  S.V.  Hering,  D.R.  Lawson,   et  al,  "Field  comparison  of
     methods for nitric acid," Atmos. Environ., in press (1987).
 2.  B.R.  Appel,  Y.   Tokiwa,   E.L.   Kothny,   "Sampling  of
     particles in the atmosphere," Atmos, Environ f 17: 1787-1796
 3.  D.J.  Eatough,  V.F.  White,   L.D.  Hansen,  N.L.  Eatough,  J.L.
     "Identification of gas phase dimethyl  sulfate and monomethyl
     acid in the  Los Angeles  atmosphere," Envir.  Sci.  Technol.  9:
     (1986).
 4.  D.J.  Eatough,  M.  Brutsch,  L.  Lewis,  L.D.  Hansen,  E.A.  Lewi?. tfc«
     Eatough,   R.J.   Farber,  "Diffusion  denuder  sampling  systems
     collection of atmospheric organic compounds," Transactions af-
     International Specialty  Conference. Visibility  Protect
     and  Policy  Aspects.   Grand  Teton  National  Park,  Wyoming,
     (1986).                                                          v
 5.  D.J. Eatough, C.L. Benner,  J.M. Bayona, P.M.  Caka,  H.  Tang,  L>
     J.D. Lamb. M.L.  Lee,  E.A.  Lewis,  L.D.  Hansen,   "Sampling f°r g*tt, J
     particle   phase  nicotine   in  environmental  tobacco   smoke
     diffusion  denuder  and  a  passive  sampler,"  Proceedf rFifi-
     Symposium on Measurement  of Toxic  and  Related
     Pollution Control Association, 132-139 (1987).
 6.  N.D. Johnson,  S.C.  Barton,  G.H.S.  Thomas,  D.A. Lane,  W.H.
     W.H. ,   "Field evaluation  of  a  diffusion denuder based
     sampler for  chlorinated  organic  compounds , "  Proceedings o£
     Annual Meeting  of  the Air  Pollution Control  Association.  24-2
     San Francisco,  CA,  Paper 84-32.1 (1986).
 7.  S,   Bertoni,  A.  Febo,  C.  Perrino,  M.   Possanzini,  "Annuls*
     diffusion  sampler:    a  new  device  for  the  collection of
     vapors,"  Annali di Chimica 74: 97-104 (1984).


                                   744
-------
^•L.  Brookes,  "A cryogenic trap and impinger for  sampling acid,  base,
°r water soluble organic  compounds in  air,"  J.  Assoc.  Pollut.  Analysts
22:  103-109  (1984).
J-D.   Pleil,   W.A.   McClenny,   "Temperature -dependent   collection
efficiency  of   a cryogenic  trap  for  trace -level  volatile  organic
impounds , "   Proceedings   of  the   77th  Annual  Meeting  of  the  Air
Dilution Control Association.  24-29  June,  San Francisco, CA,  Paper
84-17.5 (1984).
G-B.   Oldaker   III,   F.C.   Conrad  Jr.,   "Estimation   of  effect  of
environmental tobacco smoke on air quality within  passenger cabins of
Commercial aircraft," Environ.  Sci.  Technol.  21:  994-999 (1987).
"• Muramatsu,  S. Umemura,  T.  Okada,  H.  Tomita,  "Estimation of personal
6Xposure to  tobacco smoke  with a newly developed nicotine  personal
Jonitor," Environ. Res.  35:  218-227  (1984).
*-G.  Lewis,  J.D. Mullk,  R.W.  Coutant,  G.W.  Wooten,   C.R.  McMIllln,
'Thermally desorbably passive device for  volatile organic chemicals in
^bient air," Analvt. Chem.  57:  214-219 (1985).
 •K.  Hammond, B.P.  Leaderer, "A diffusion monitor  to  measure  exposure
^° passive  smoking," Environmental Science  &  Technology  21:  494-497
(1987).
S-K-  Hammond,  J. Coghlin,  "Field study of passive smoking  exposure
^ith passive  sampler,"  Indoor Air  '87.  Proceedings  of  the  Fourth
fei££rnational Conference  on Indoor A|r  Quality  and Climate.  Berlin
'West),  17-21 August 1987,  Seifert B.,  Esdorn H. , Fischer M. ,  Ruden
!?•.  Wegner J.,  eds. ,  Institute  for  Water,  Soil and Air Hygiene,  Vol.
J. 131-136 (1987).
 Environmental  Tobacco Smoke.   Measuring  Exposure and Assessing Health
Infects." National Academy of Sciences, Washington, DC (1986).
 The Health Consequences  of Involuntary Smoking," U.S.  Department of
J6alth  and Human Services (1986).
Q-J.  Eatough,  C.L. Benner,  H.  Tang, V.  Landon,  G.  Richards, F.M.  Caka,
 • Crawford,  E.A.   Lewis,  L.D. Hansen,  N.L.  Eatough,   "The  chemical
c°mposition   of   environmental  tobacco smoke   III.  Identification  of
c°nservative tracers of environmental  tobacco smoke,"  Environ. Inter..
      te.
 •w-   Ogden,   K.C.  Maiolo,   "Gas   chromatographic   determination  of
s°lanesol  in environmental  tobacco  smoke (ETS)," J. High Res.  Chrom.
^Q£brom.  Comm..  in press  (1988).
j-L-  Benner,  J.M. Bayona,  F.M.  Caka,  H.  Tang, L. Lewis, J.  Crawford,
 •D-  Lamb,  M.L.   Lee,  E.A.  Lewis,  L.D.  Hansen, D.J.  Eatough,  "The
c
         composition of  environmental  tobacco smoke.  II.  Partlculate
jjliase,»   Environ.  Sci.  Teehnol..  submitted (1988).
 •K-  Hammond, B.P. Leaderer, A.C. Roche, M.  Schenker,  "Collection and
*nalysis  of  nicotine as  a marker  for  environmental  tobacco  smoke,"
frgs^ Environ.  21:  457-462 (1987).
;"'W-  Eudy, F.A,  Thome, D.L. Heavner,  C.R.  Green, B.J.  Ingebrethsen,
 Studies  on the vapor-particulate phase distribution of environmental
?lc°tine  by  selective  trapping  and detection methods,"  Proceedings^
ffilL-Annual  Meeting of  the A?r  Pollut.  Contr.  Assoc..  22-27  June,
^neapolis,  MN,  Paper  86-38.7  (1986).
jlj-  Eatough, C.L. Benner,  J.M.  Bayona,  F.M. Caka,  G.  Richards,  J.D.
r^b,  E.A. Lewis, L.D.  Hansen, "Chemical composition of environmental
 °bacco  smoke.  I. Gas phase acids and bases," Environ.  Sci.  Technol.,
and
      ted       .
 'J<   Eatough,  E.A.  Lewis,  C,   Benner and  N.L.  Eatough,  "Gas
 «rticie  phase nicotine  In environmental tobacco  smoke,"  Proceedings
    he_ 79th Air  Pollution Control Association  Meeting.  22-27  June,
    eapolis,  MN,  Paper No.  86-68.5 (1986).

                               745
-------
                                                                        i"
24.  M.W.   Ogden,   "Gas  chromatographic   determination  of  nicotin* ^
     environmental  tobacco  smoke:   collaborative  study,"  .T,   Assofij——
     Anal.  Chem.. submitted  (1988).                                     efll
25.  F.A.  Thome, D.L.  Heavner,  B.J. Ingebrethsen, L.W.  Eudy,  C.R- ^^e
     "Environmental  tobacco  smoke monitoring with an atmospheric  preSa
     chemical  ionization mass spectrometer/mass spectrometer  coupled   ^
     test  chamber,"  Proceedings  of  the 79th ^Annual Meeting
     Pollut.  Contr.  Assoc.. 22-27  June,   Minneapolis,  MN,  Paper
     (1986).
Table  I.  Concentrations  of  Gas  Phase  Nicotine  in  Environmental
Smoke in a 30-m3 Teflon Chamber, nmol/m3, Determined With Several
Sampling Methods.
# of
Gig.
4
4
4
4
1/2
1/2

1/2
1/2
1/2
Sampling Annular Reynolds BYU XAD-II/
Time, rain Denuder XAD-IV XAD-IV Filter
94
120
15
15
120
120

15
15
15
53251102 53771 31 52541 81
39071143 37841 57 35941320 36531760
57621133 64961 76 54871 88
36931652 51931143 37861516 42101314
12671135 6201 0 511± 19 361
6241 1 6351 10 522 5571106

10271 18 8981131
10231192 11021 12 1273
8821 70 9751 9 11651150
Filter/ P^gj
artrf1
480*
0
29251274 325'"
1064* l3
jl
41411121 I122'
552169 l35
656123
af>9
yu'
s2,Q
S*$
                                   746
-------
Tab]
 b        Concentrations  of  Particulate  Phase Nicotine  in Environmental

   CC° Smoke in a  SO-""3  Teflon  Chamber,  nmol/m3,  Determined With Several
Dtff
   •
  •etent  Sampling Methods.  The Sample Sequence is the Same as in Table I.
"i.
vl-6*.
4

4

4

4

1/2
' t.
1/2
' «.
1/2

1/2

1/2

Campling
Time . min

94

120

15

15

120

120

15

15

15
Annular
Denuder

243.6+2.5

234. ±45

271. ±64

101. ±57

17.6+0.5

63.0+2.3



30.9±0.1

32.5+2.1
Reynolds
XAD-IV

14 . 5±0 . 2

9.9±3.6

3.3+3.3

14 .4+3.2

0 . 0+0 . 0

2.0+0.0

0.0+0.0

0.0±0.0

2.0+2.0
XAD-II/
Filter



29. +27



22. ±18

12.1+1.4

6.2+5.4

14.2±0.9

6.5

8.3+0.4
Filter/
XAD-II



113. ±25



78.8±1.0

30.6±6.3

66.8+1.7

66.7±9.1




                 Gas Phase
                                            Particles

  '      Gas and particulate phase distribution of alkaloid compounds  in

           tobacco smoke in a  30-m3  Teflon chamber.
                                747
-------


CO
E
"o
c
L_
C

D~

CO

•S
 ° A

     o
     o

    z
                A

            Denuder
         7000
                            o

                        XAD/Filter
     D

Filter/XAD
Slope »  1
              0     1000   2000   3000   4000   5000   6000



                       [Nicotine(g)]  XAD-IV,  nmol/m3



Figure  3.     Comparison  of  the   concentrations  of gas  phase

determined  with  annular diffusion  denuder,  XAD-II and  XAD-IV

systems.  Each data point is the average of replicate samples.
                                                              7000
                                  748
-------
       #  Cig,   4    4    1/2    4     4   1/2   1/2   1/2   1/2
  Sample, min  120  120   120  is.o  15.0  15.0  15.0   15.0  15.0
        7000
   n
    E
    r
    c
    O
    0


                 Denuder
                        XAD-IV
                                                  PSD
         Comparison  of  Che   concentrations   of   gas   phase   nicotine
      with an annular diffusion denuder,  an XAD-IV sampling system and
      diffusion sampler,  PSD.   Each  data  point  is  the  average  of
c&te  samples.
   #  Cig.   4     4
Time, min  120   120
                             4
                           is.o
                         4
                        15.0
1/2
120
1/2
120
1/2
15.0
1/2
15.0
          300
    o
    c
   I
    
-------
     Results  from  the  Environmental  Response
Preliminary  Evaluation  of a  Direct  Air  Sampling
Spectrometer  (the Bruker MM-1)

Robert E. Hague
Department of Environmental Science
Cook College, Rutgers University
New Brunswick, N.J.  08903

Thomas H. Prichett
U.S. EPA Environmental Response  Team
GSA Raritan Depot
Edison, N.J. 08837

Kwong Cho
Roy F. Weston, Inc. (REAC)
GSA Raritan Depot
Edison, N.J. 08837

Ben Shapiro, formerly of
Envi response, Inc.  (EERU)
GSA Raritan Depot
Edison, N.J. 08837

     During the summer of 1987,  an investigation of the
MM-1  mobile  mass spectrometer  was  performed  in
evaluate that  instrument's  applicability  for
response  and Super fund-related  tasks.  As a  tech
application of mobile,  direct air sampling mass  spect i
to environmental  applications is  fairly  recent,  and/13
potential of assuming  a role  of major  importance  i*1
toxics measurement.   The  evaluation  was  divided into
phases.     The  first  phase   evaluated  the  instr7e<3
sensitivity, linearity and reproducibility for a  sele
of organic compounds.   The second phase  of  the study e  fe$
the  instrument's  accuracy when challanged with   v*ixt**
known target compounds.   The  third phase challenged
with  a  series of unknowns containing  mixtures  of
known to present  analytical problems to  conventional
methods.   Conclusions are drawn  concerning  the use
monitoring of direct air sampling mass spectrometry.
                            750
-------
 8p
  gription of the  Instrument
ijw ne MM-l  is  a full  quadropole  electron  impact  mass
!>-[ trometer which  has been  designed  specifically  for  use
 !af field conditions.  The unit  is easily mounted in a four-
    drive vehicle, and is rugged enough to allow dependable
         at  off-road sampling points.    During  mobile
        , power is supplied  using a  rechargeable 24 volt DC

is . Operation  is on a real-time analysis format.   The sample
jjjj rawn into the instrument at  a  constant rate in ambient air
&6p  ^halyzed  instantaneously  and  continuously.  There is  no
fyntJLat*on of    compounds  or  sample  preparation,  and  all
          are   analyzed  simultaneously.     This  creates
           in compound identification.   Analytical  results,
         parameters,  and  listings of identified compunds are
    Yed on a  video screen and continuously updated.   During
   ne air sampling,  ion  peak intensities  are  displayed  as
   ~rams  whose  height  varies as  the base 10 logarithm of the
 ,r  intensity.   The  intensity  scale range allows measurement
 t  ' orders of  magnitude.   All results are  normalized,  so
   Direct comparison of measurements from  all  analyses can
         Printed copies of  analyses,   library  contents  and
     status may be had at any  time via  a small panel-mounted
J>H»>jSan»Pling  is  performed using  a  sampling  line  directly
7 a e   The samPle head  is  a  nickle  gauze  coated  with  a
    rineable  silicone membrane.   Samples are  pulled  through
   ^embrane  by  a sampling pump at  1  to 3 cc/minute.   The
  i ne  serves to protect  the  sample line  itself, which
     s  of a  3-5  meter 0.32 mm quartz capillary column coated
      SE-54  phase within  an insulated,  heated jacket.   At
      of  this  line, the sample is  drawn  through a  second
       membrane  and into the  mass  spectrometer.   in  the
   -°f  line  contamination, there is a backflush system.   The
     g  line is  capable of temperature ramping and  some
      e  separation of  compounds can be achieved.    Once
     the mass  spectrometer,  the  compounds  are ionized
     ng  electron bombardment by an electron source.  (Figure

  flln general, gas chromatographic/mass spectrometer  systems
    Si9ned to present  individual  molecular species to  the
     ectrometer  following their separation from  the  sample
     on the  chromatographic  column.  In  real  time analysis,
       is the case  with the  MM-l,  the compounds are  not
        and  are  presented  to  the detector  simultaneously.
        from  different  compounds  are combined  and  the
        ion  mass ratio is the  sum of  the  ions from all of
   o   ounds   present in  the sample.    This  leads  to  the
,   °usiy mentioned difficulties in  compound  identification
 WJantification.   In most  environmental  analyses  of
         the three most  prevalent  characteristic  ion  masses
                            751
-------
of each compound  are  utilized in identification.  In an effort
to circumvent  the problems encountered  in  real-time analysis
with identification, the the MM-l utilizes four characteristic
ion masses  and their  abundance  ratios.  The  selected ions and
their relative  abundances may be entered by  the operator from
the instrument  keyboard.   If  a pure sample  of the compound is
available,  a  spectrum  may be added  to memory  by taking a
direct   sample  of  the   vapor,   selecting   up  to   four
characteristic  ions from the  resulting  spectrum   and saving
the named compound  to memory.  The MM-l has been provided with
routines  for  the  avoidance  of  false alarms.    One  or  two
"impossible"  ions  (i.e.  ions contributed  by  an interfering
compound) may  be  entered with an abundance  of zero.   If the
ions of the target compound and  the  impossible  ion appear at
the  same time, with  the  impossible ion  at  some  intensity
greater than zero, the  alarm signal  for the target compound
will be  surpressed until  there  is  an excess of  the  desired
compound over  the interferant.    Even  using  this  approach, it
was  found  that  with  complex  mixtures of compounds,  the
detector  has  difficulties with identification, and  false
positive and false  negatives are  common.

Experimental

     In  order to  evaluate the  MM-l  over   a  range  of
concentrations  representative  of  both  chronic and  acute
emission  levels of organic  compounds in air,  a  dual sample
dilution manifold was constructed to provide the capability of
diluting cylinder gas standards  in ambient air over the ppb to
ppm  range.   The system  consists  of  two  parallel  sample
dilution lines  of differing diameters with  side-tapped glass
and Teflon  sampling ports.   Both manifolds  use   individually
sized mass  flow controllers at their discharge  ends and share
common diluant sources,  (ambient air and  zero  air)  common
organic vapor  sources and  a  common  low  pressure multi-stage
blower.   The manifold lines were  sized so as to provide usable
concentrations  over a feasable sampling interval.  The systems
consisted specifically  of  a 7/8  inch ID Teflon  line  for the
high dilution  line, providing metered flow values  of 30-150
liters/min.  for concentrations of 5  to 100 ppb  and a 3/8 inch
ID stainless steel  line providing metered flow  rats of 2-7,5
liters/min.  for vapor concentrations over the  range from 1.0
to 10.0 ppm.  Both  lines were heat traced and maintained at 40
degrees C.  Dilution ratios and  flows are shown in Table 1.

     The shared vapor sources were of two types:

          Standardized multi-componant compressed gas mixtures
that were  metered  through a  mass  flow  controller into  the
sample lines for  ambient air dilution to produce the  desired
concentration and,

          A heated vaporization  source for compounds which are
unsuitable  as  pressurized cylinder  standards.  This consisted
of a  series of midget  glass impingers  containing the  neat
                             752
-------
 compound  of interest  kept in  a constant  temperature  water
 bath.  A controlled air flow is saturated with the vapor as it
 passes through  the impinger  train  and diluted  to its  final
 desired concentration.
     The  evaluation of  the  MM-l  was  divided up  into  four
 phases.   Phase  1  evaluated  the instrument's  linearity  and
 sensitivity by providing progessively higher  dilutions  of
 standard  gas mixtures cylinder by cylinder over  the  test
 range, and a "detectablity limit"  for  those compounds in that
 mixture determined.    Phase  2 presented  the instrument  with
 unknown concentrations of  known  compounds within the  list  of
 compounds presented to in Phase 1.   Phase 3  presented  the MM-l
 with mixtures of compounds from a list  of non-cylinder-stable
 compounds.  The analyst was  not  provided with information  on
 compounds to be expected  or their concentration.   A  list  of
 the compounds present  in each cylinder  is given in Table 2.

 Phase 1 Results

     Phase 1 testing  indicated that the overall  sensitivity
 was  very  much  a  function  of the  compound mixture under
 consideration.   Detection  limits  for  some  compounds were found
 to be as  low as  10 ppb, depending on the compound.  It  should
 be emphasized that the detectability  of  a  given compound was a
 function  both of the other  compounds present and the  ambient
 background  concentration of organic  materials.    Complex
 mixtures create  interferences  which raise  the  detection limits
 considerably.   (Table  3.)  Common-ion   effects  apparently
 prevented  the MM-l from recognizing some compounds even at the
 10 ppm level.  However, it should also  be emphasized that the
 linearity  of response  for many  of  the compounds was  quite
 good, with correlation coefficients  exceeding  0.9 over  three
 orders of magnitude.   Typical response  curves are  shown in
 Figures 2  to  5.

 Phase 2 Results

     The results of  tests performed are indicated in  Tables 3
 and  4.    As  may  be  seen,  in  complex  mixtures both  false
 positives  and negatives  were  common.    For  those  compounds
 which were  correctly   identified,  quantitation  error  ranged
 from  +2.2% to -39.1% and was dependant on the compound.   The
 importance of the  complexity of the  compound  mix  becomes
 readily apparent,   if   the  Phase   1   limits  of detection are
 compared  with  the false  negatives  in Unknowns 1  to  4.
 Reductions in sensitivity by factors of five to twenty or more
 occurred for individual compoundsin each of the unknowns.
 (Tables 4 and 5)


Phase  3 Results

     Phase 3 presented a  series of mixtures to the instrument
whose concentration and composition were unknown
to  the  analyst. An effort  was made to select  compounds which
                             753
-------
are  of interest at  Superfund  sites,  but  which were   e
available as cylinder gas standards.  Calibration curves *
not prepared for these compounds  and the emphasis was  stric ^
on qualitative  analysis.   The results  are  shown  in Tab*6 Of
The  MM-1  demonstrated  its  ability  to  identify a  numbe*^^
compounds (cresol, pyridine) which have been known to
analytical  difficulties to  gas  chromatographic

Overall Comments

     o    Sensitivity and Accuracy

     For scenarios where individual compounds or a mixture ^
compounds with  no common  ions  the  sensitivity of  the *"
under  typical  ambient  conditions  lies  in  the  10 to 25
range.  The  stability of the  instrument  is such that
readings over a period of time  agree within  a  fraction
response unit.   Linearity  of  response  was found  to  l>e *
over the range  of 10 ppb to  10  ppm for  many compounds.   fe
stated previously, mixtures of compounds sharing one or  ^
common ions drastically reduces both  the sensitivity  ^
accuracy of the instrument.   Accuracy was also found to  ^^6
where  compounds  not  included in the  target  library
present in the  mixture,  which  may  well  be the case at
sites.
     o    Portability                                   _.,flted
     Subsequent to the  Phase  3 testing, the MM-1 was
in a four wheel drive vehicle  and  field tested.  It was
that the durablity of the unit  was  truly remarkable.
hours  of off the  road use,  the  instrument still retained
calibration and  full  internal vacuum.    The battery
found to generate sufficient power for  a 6 to  8  hour
schedule before recharging was necessary.

     Setup  time  is  minimal   at about  15 minutes
additional 30 minutes for calibration and sample line
The analysis time is  approximately 15 seconds, with trip
analyses in less than one minute.

     o    Level of Operator Training Required

     The  Bruker  MM-1  is  designed  with  extremely
operating procedures  based  on a  series menus  all acces
from a simple  keyboard.  This allows  an operator to a*1
air and soil samples  with a minimum of  training. It snoUA
noted  however, that although the  instrument is  siinp1^
operate, the data have  limitations and should  be assesse ^
and  experienced  mass  spectrometer operator  aware °*
interferences which are inherent  in real-time  analysis.

     The experienced operator will be able to recognize
positives and false negatives  by reviewing the  ion was-  -1J1;,
and the abundances.   The opeator  should have and understan ^j,
of the theory behind the operation of the  instrument  as w  -»"
With this understanding,  an operator will be better
                             754
-------
  rfccognize any problems which may occur.

   o    Reliability

$ta The  reliability of  the  instrument from  a a mechanical
Sth  °int was  unquestionable.   Over  the course  of three
W s'  tne sole  breakdowns  were a  leaking  calibration gas
ing e  and an electronics overload caused  by  a power  failure
boj.. Susfcquent voltage surge.   The manufacturer was prompt in
fW cases  and  in both cases  the  unit was  functioning
  Iectly within one day.

   °    Observations and Recommendations

Is  In the  configuration used  in this study, the instrument
W ot  applicable  to  most  site  assessment  work.    The
thj ruroent is designed for air monitoring and  performs well in
fiw capacity.   The complex mixtures  of compounds which can be
^Ohi  *n waste  sites are  likely to  create   identification
  °lems with false negatives  and  positives occur ing.
tiie In the case of an emergency spill  the  instrument would be
lts Method  of  choice in delivering  quick  and precise data.
»^Q transportability and ruggedness would allow a  plume of
   r to readily traced  and quantitated.
      relatively  new  gas chromatographic  attachment  is now
           Although  it was  not evaluated  in  this  study,
    .cturer's  data indicate that  many of  the  problems
     fied  in  this  study  should   be   resolved  by  this
     ment.
w°ve
                      REFERENCES
    Bruker-Franzen  Analytic  GmbH,     The  MM-l  Mass
    Spectrometer  User  Manual.   Bruker-Franzen Analytik
    GmbH, Bremen, West Germany, 1986  2-28 pp.
                          755
-------
Table 1.  Dilution Ratios,  Sample  Generating Mani*0
          Cylinder Gas Dilutions
Required
Concentration
10 ppb
25 ppb
100 ppb
1000 ppb
10000 ppb
Dilution
Ratio
5000/1
2000/1
500/1
50/1
5/1
Flow
Rates
150 slpm
30 sccpm
60 slpm
30 sccpm
7.5 Slpm
15 sccpm
4.0 slpm
80 sccpm
4.0 Slpm
800 sccpm
Dilution
Line
7/8 incn
7/8 iHC»
3/8 in°n
3/8 i*°n
3/8 il*cn
                        756
-------
     2.        CYLINDER  GAS  STANDARDS—PHASES  i and  2
COMPOUND             CONCENTRATION  (PPM)
TOLUENE                        46.7
1,1,1-TRICHLOROETANE          47.48
1,4-DIOXANE                    42.82
ACETONE                        50.05
1 , 2-DICHLOROETHANE             47 . 67

B_
COMPOUND             CONCENTRATION  ( PPM )

VINYL CHLORIDE                 49.9
BENZENE                        49.98
METHYLENE CHLORIDE             50.16
1 , 1-DICHLOROETHYLENE          46 . 75
TRICHLOROETHYLENE              53,15

C
COMPOUND             CONCENTRATION  (PPM)
METHYL ETHYL KETONE            46.05
HEXANE                         58.86
METHYL ISOBUTYL KETONE         48.69
TETRACHLOROETH YLENE            53.05
1,4-DIOXANE                    48.10
COMPOUND            CONCENTRATION  (PPM)
CYCLOPENTANE                  48.63
ETHYL ACETATE                 49.91
1,1-DICHLOROETHANE            49.01
1,1,2-TRICHLOROETHANE         51.08
CARBON TETRACHLORIDE          50.19

£

COMPOUND            CONCENTRATION  (PPM)

CHLOROBENZENE                 26.13
0-CHLOROTOLUENE               24.36

E
COMPOUND            CQNCENTRATION  (PPM)
ISOPROPANOL                   48.47
ETHYL ETHER                   48.65
3-CHLOROPROPENE               49.09
STYRENE                       55.10
ETHYL BENZENE                 50.77
FREON 11                      51.05
                        757
-------
Table 3.  Phase 1  MM-l Limits of Detection
     Compound
     Acetone
     Vinyl Chloride
     Cyclopentane
     Benzene
     Methylene Chloride
     Hexane
     Ethyl Acetate
     1,4-Dioxane
     Toluene
     l,1-Dichloroethylene
     1,1-Dichloroethane
     Ethyl Benzene
     Chlorobenzene
     o-Chlorotoluene
     Trichloroethylene
     1,1,l-Trichloroethane
     1,1,2-Trichloroethane
     Tetrachloroethylene
     Styrene
     Trichlorofluoromethane
     Ethyl Ether
     Methyl Isobutyl Ketone
     1,2-Dichloroethane
     Methyl Ethyl Ketone
     Isopropanol
     3-Chloropropene
     Carbon Tetrachloride
Limit of De.tefitA.QJlj ppbv
          100
         1000
           10
           25
           10
           25
           10
         1000
           25
           25
          100
            5
           10
           10
           25
          100
           25
           25
          100
          ND >10 ppm
         1000
        10000
          100
          ND >10 ppm
          ND >10 ppm
          ND >10 ppm
          ND >10 ppm
                            758
-------
Table  4.  Phase 2  Mixtures
                      vVMC *
Fain
 Hog.      Pos.
Acetone
1,4-Dleiane
l,t Olchloroethane
1,1.1 Irtchloroethane
Toluene
Chlorobenione
o>Chlorotolitene
Acetonltrlle
Dlchloreaethane
Acetaldehyde


Compounds
lenxtne
Vinyl Chloride
Methylene Chloride
l,l-D1chloroethylene
Trlchloroethylene
Ethyl lenzene
Ethyl Ether
Freon-H
Styrene
Acetonltrlle
Allyl Chloride


Conpounds
Methyl Ethyl Ketone
Hexane
1,4-Dioxane
1,1 Dlchloroethane
1.1.2 Trlchloroethane
Cyclopentane
Cthyl Acetate
Carbon Tetraehlorlde
Tetrachloroethylene
Acetonltrlle
Methyl Isobutyl Ketone
MO. «
421. "
476. x
474. *
467. I
121. X
127. x
x
X
X
MM Response: Unknown 12
Cone. MM False
(»pb) Identified Neg. Pos.
133.7 x
133.7 »
134.4 x
125.3
142.4
264.7
249.3
25S.3
279.2 x
x
249.7 *
MM-1 Response: Unknown 13
Cone. F«l»* f*1**
(ppb) Identified Neg. Pos.
124.0 "
158. 5 x
129.5 x
247.0 x
257.5 x
245.1 *
251.5 x
253.0 x
143.2 x
x
131.1 +
                        detected - not quantified
                       m-\ Response:  Unknown 14
Compounds
Vinyl Chloride
lenzenc
Hethylene Chloride
l,l-01chloroethylene
Trlchloroethylene
Cyclopentane
Cthyl Acetate
1,1-Dtchloroethane
1,1,2-THchloroethane
Carbon Tetraehlorlde
Cone.
(PPb)
251.3
251.4
252.8
235.6
267.9
564.1
579.0
568.5
592. $
582.2
Identified
x



x
X

X
X

False
Neg.

x
X
X


X


X
False
Pos.










                            759
-------
        Table 5.   Accuracy of Phase 2 Responses

                     Generated            Measured             Pcrcc?ofi
Compound           Concentration (ppb)   Concentration (ppb)   Deviate
Chlorobenzene
Toluene
o-Chlorotoluene
      522
      468
      527
                                            321
                                            507
                                            540
Compound
Accuracy of MH-1 Response:  Unknown 12

    Generated            Measured
  Concentration  (ppb)  Concentration  (ppb)  Dev
Benzene
Dichloromethane
Styrene
        134
        134
        279
                                               137
                                               155
                                               170
Compound
Accuracy of HM-1 Response:  Unknown 13
                                               ..
    Generated             Measured              .fa
  Concentration (ppb)   Concentration  (ppb)   Devi*
                                                                   ll011
1,1,2 Trlchloroethane
Ethyl Acetate
1,1 Dlchloroethane
Methyl Isobutyl Ketone
                       NQ -
          258                155
          252                219
          247                229
          131                NQ
          detected,  not quantified
                                                              -39.J
                                                              -13.1
                                                              -  7.3
                   Accuracy of MH-1 Response:   Unknown 14
Compound
                         Generated
                       Concentration
Vinyl Chloride
Cyclopentane
1,1-Dlchloroethane
Trichloroethene
1,1,2-Trichloroethane
                             251
                             564
                             569
                             268
                             593
                           Measured
                         Concentration
                             170
                             607
                             555
                             165
                             532
                                                                  <
                                                              - l\
                                                              -38.J
                                                              -10-3
                                  760
-------
      6.  Phase  3  Compound  Identification

 Unknown #1
Aniline
^-Cresoi
pyridine
Ethoxyethanol
^cetaldehyde

Unknown #2
               Concn.
                (ppm)
 1.34
 0.5
 45.0
  7.0
               Concn,
               (ppm)
                          False     False
              Identified  Positive  Negative
                          False     False
              Identified  Positive  Negative
 ~ Cresol
Jcetaidehyde
£thyi Benzene
 0.5
79.0
 6.7
^known #3
               Concn.
               (ppm)
                          False     False
              Identified  Positive  Negative
Chloroform     7.8
JJutyr aldehyde  4.6
               6.0
     oromethane —
                          761
-------
Figure 1.  Schematic of the MM-1 Inlet System
                        762
-------
  FIGURE 2.  METHYLENE CHLORIDE
      100
                       1000
                                       10000
            CONCENTRATION (PPB)




FIGURE 3.  1,1,2-TRIGHLOROETHANE
                                        toooo
-------
                   FIGURE 4.  BENZENE
0
§
LJ
                        CONCENTRATION (PPB)
               FIGURE 5.   CHLOROBENZENE
§
111
£
 I
       10
100


 CONCENTRATION (PPB)


   764
                                      1000
-------
        °F GAS PHASE RETENTION VOLUME BEHAVIOR OF
         C°MPOUNDS ON TENAX-GC AND OTHER SORBENT
 0*;epS°n Graduate Center
     '"nent of Environmental  Science and Engineering
      ^•W. Von Neumann Drive
    erton, Oregon 97006
     There is a need for to able to predict  the  compound-dependent volumes
   gas that can be sampled with little breakthrough with the sorbent Tenax-
  '   Application of linear Brunauer-Emmett-Teller  (BET)  isotherm principles
 ^ atgs the retention volume per gram of sorbent at temperature T (K)
 jj s,T> L/g) to the pure compound vapor pressure at  T (p°,  torr).  Trouton's
 b    transforms  this equation into one between log  Vg
-------
 Introduction

      Over the past 15 years, there has been great interest in the use
 Tenax-GC as a sorbent for sampling gaseous organic compounds.  This    .
 material has been employed in the collection of organic compounds in   ,c4J
 ambient and workplace air.  It has also been utilized in certain analy ^
 methods such as purge and trap (P&T) .  (The P&T method is applied routin
 in the analysis of water samples for volatile organic compounds.)  AS a
 result of this broadly-based interest in Tenax-GC, considerable data »a
 been gathered over the years on the gas phase retention volume values
 various organic compounds on this sorbent.

      Until the recent work of Pankow1 ,  the data on Tenax-GC has  not bee ^
 reviewed,  compiled,  and analyzed in a manner that maximizes its utili^. flf
 scientists interested in sampling gas phase organic compounds.   The v°
 Pankow  provides a review of the  available retention volume informati0
 on Tenax-GC,  and also examines the theoretical basis for expecting *e  s
 tion volume values to correlate with physical constant parameters sue n ^
 boiling point and vapor pressure.   Specific correlation equations base
 various published retention volume data sets were presented in that w°
 This paper provides  an overview of that work and also considers  the o3
 available  for polyure thane foam (PUF) .

 Theory

                              Retention Volume

      The specific gas phase  retention volume of a compound (V j,
 sorbent at a  given temperature  (degrees  Kelvin)  is given by tne  equa

                      Vg,T - Vw  -   cs/cg

 where:     V^   -  net gas phase retention  volume (L) ;
           w    -  weight  of  sorbent  (g) ;
           c    -  sorbent phase concentration (mol/g) ;  and
           c    —  gas phase  concentration  (mol/L) .

      The dependence of V   T on temperature  is a  function  of the
 desorption AHS (kcal/molj!  Chromatography  theory  predicts  that
                                                                     (2)
               d log  (Vg(T/T)/d(l/T)   -    AHS/2.303R

where:    R   -   gas  constant (0.00199 kcal/mol) ;  and
          T   -   temperature (degrees Kelvin) .
                                                                       tli
      Equation 2  is very useful because it allows one to:  1) determi
retention volume of a given compound at a temperature sufficiently " &
that  the compound moves through the sorbent bed at a measureable
2) extrapolate down to the temperature of actual interest.
                                                                        ^
When sorption is taking place to a limited number of sites, the
                              er.
                              9  •?
said to be "Langmuirian" in character.  For site-limited sorption, w
                 c   -
                                   766
-------
  AIs°-  by the ideal gas law,

                  cg  - p/760RT                                       (4)
  Ilk
    re:     6    —  fraction of surface sites occupied (dimensionless);
           NS   -  moles of sites/cm  (mol/cra );  and
           A    -  specific surface  area of sorbent (m /g).
  On>bining equations  1,  3,  and  4, we  obtain
            Vg,T  "  cs/cg  "   *NSA<104 cm2/ni2)  760RT/p               (5)

         gas/solid sorption at  low surface coverages,  the  linearized
        of the Langrauir adsorption isotherm is

                       0   -   bp/760                                 (6)

   r&:    b   -  Langmuir constant (torr  ) ;
          p   -  gas phase pressure of compound  (torr) .


                    Correlation of log V  T with log p£

 p    The Brunauer,  Emmmett and Teller (BET) theory of gas/solid sorption
  edicts that
                         exp [(AH  -
                b  -  f  ....................                        (7)
                               p°/760

               -  constant related to entropy of desorption
                  (dimensionless) ;
               -  enthalpy of vaporization of pure  liquid (kcal/mol) ;  and
               -  vapor pressure  (torr)  of pure  compound at temperature T.

    At T - 293  K (20  °C) ,  a  combination of equations  5   6,  and  7  together
    the assumptions that:  1)  f =_l'  2) Ng   - 6  x  10 "10 mol/cm2  (Pankow,
    ! and 3) A  for Tenax-GC  -  6.4 m2/g  (Pankow1), yields

         l°gVgi293  -   -  0.15  +   0.74(AHS  - Aty  - log  p£93      (8)

%  ^° t'le extent that the parameter 0.74(AHg - AH^) remains constant  from
fl^t  Unt* to compound, equation 8 reveals that a correlation of log V  293
V^, Vs<  ^°S P293 snould yield a straight line with a slope of -1.00.'
*te 6 tlle actual slopes found by Pankow1  are different  from -1.00,  they
- 0 ^^negative and relatively close to -1.00.  For PUF, the value of A is
tiev' 2m /g, but  the corresponding theoretical  equation  for that sorbent will
      e^ess  be  similar to equation 8.  Some actual correlations based on
          data  are given  in Table 1 for both Tenax-GC  and PUF.
                   Correlation of log V  T with Tb (K)

        integrated  form  of  the Glausius-Clapeyron equation is

                log p°,  -   (log  760  -  B)Tb/T   +  B                  (9)
                                 767
-------
where:    Tb - boiling point  of  the pure  compound  (K)  - boiling point
               +  273
          B  - a  compound  dependent constant  (~  7.7)
     At T - 293 K  (20  °C),  a  combination of  equations  8  and  9  together
with the assumption  that  B  -  7.7 yields  (Pankow,  1988)
log
                       -  -7.9 + 0.74(AH
                                                 + 0.016 Tb
(10)
     To the extent that the parameter 0.74 (AH
                                                       remains  constant
from compound to compound, equation  10   reveals  that  a  correlation of
V
   2
Wfiil
    93 data vs- Tb  (R)  should yield a  straight  line with  a  slope  of
                                                                      -° '
    tj *j           U            •+         1
    e the actual slopes  found by Pankow for Tenax-GC are somewhat
different from 0.016,  they are all at  least moderately  close  to  that
value.  Some actual correlations are given in Table  1.

Table 1.  Examples of  Correlation Equations for Gas  Retention Volume
          on Tenax-GC  and PUF at 20 °C  (293 K) as Functions of log P293
          (torr) and Tb  (K).

compound   sorbent     log V  293 vs log p293     log V 293  vs  Tb
 tvPe                  	:	^   	-:	o'*,
slope intercept
mixed
alcohols
amines
PAHs
Tenax-GC
Tenax-GC
Tenax-GC
PUF
-1
-1
-1
-1
.44
.42
.00
.10
4.
2.
3.
1.
30
73
13
33
rz
0.
0.
0.
0.
slope intercept f
96
99
99
99
0.0258 -7.43
0.0369 -12.59
0.0183 -5.09
NA NA
0.9?
0.99
0.98
NA
NA - not available.
__-- ~-
Conclusions
     Existing literature data on retention volumes of organic compound5
various sorbent phases are amenable to linear regression vs. physical
constant data such as vapor pressure and boiling point.  The resulti^S ^
regression equations will be very useful in designing sampling schemes ^
gaseous organic compounds in the workplace, in the ambient atmosphere*
in the indoor environment.

References
1. J.F. Pankow, "Gas Phase Retention Volume Behavior of Organic
   on the Sorbent Poly(oxy-m-terphenyl-2',5'-ylene)",  Anal. ChenL_>
   950.(1988).

2. R.H. Brown, C.J. Purnell, "Collection and Analysis of Trace
   Vapour Pollutants in Ambient Atmospheres," J. Chromatopr. .
   (1979).

3. F. You, T. F. Bidleman, "Influence of Volatility on the Collection
   Polycyclic Aromatic Hydrocarbon Vapors with Polyurethane,
   Sci. Technol..  1£:   330. (1984)
                                   768
-------
        STUDIES CHARACTERIZING ORGANIC EMISSIONS
     UNVENTED KEROSENE SPACE HEATERS:  PHASE  II
 ?a]rlCia M" Boone and Brian p- Leaderer
 JQ, e University Department of Epidemiology and Public Health
 }je n B. Pierce Foundation Laboratory
  v Haven, Connecticut
      B.  White
 »  •  EPA, Air and Energy Engineering Research Laboratory
  ^arch  Triangle Park, North Carolina

 j/ Catherine Hammond
 jj n!'ily and Community Medicine
 55 yersitv of Massachusetts Medical Center
 ^ Lake Avenue North
   Cester, Massachusetts
Co   ^en  chamber  experiments  measured gas  and particle emissions from
(fc/B?ctive  (C) , radiant  (R) ,  convective/radiant  (C/R) , and  radiant/radiant
      kerosene heaters.   Heater type  and use  affected  the composition and
Da Unt  °^  organic  emissions more  than C09,  CO,  NO  ,  S0_,  or  inorganic
  tl:1cle emissions.                      ^        X

ju  Aerosol production was similar  for  the C,  R,  and  C/R heaters.
aer    s identified  99±14% of  the aerosol mass.  Most,  77±6%,  of  heater
ca °s°l was ammonium and  sulfate.  Carbon content  was  5-34%.   Elemental
H0 °n accounted for 3.5-30% of the  carbon;  C and  C/R  heaters  produced
eiate elemental carbon than the R and R/R heaters.  Organic aerosol
   SSi°ns  ^yS/S fuel) were higher for the R/R and  C/R  heaters  than  for the
    3C heaters-  Non-volatile emissions collected by  the XAD-2 cartridge
     ^ Were hi8hest for  the R/R heater.  Volatile organic emissions were
       Co bac^ground levels for all heaters.  The highest concentrations
  ft  °bserved during heater start-up  and shutdown.  Organic compounds
C0 , cified were unburned  fuel and oxidized  combustion products.   Fewer
   Ustion products were  observed if most of the 
-------
                               Introduction

     Past chamber experiments and field studies have investigated carbon
monoxide  (CO), carbon dioxide (CO ) , nitrogen oxides (NO ), and sulfu*
oxide  (S0«) emissions from unvented kerosene space heaters and the
resulting human exposures.  Concentrations of these contaminants in
residences can exceed levels specified in health guidelines and standard
Efforts to reduce levels of these classical combustion emissions have
resulted  in new heater designs.

     Particle and organic emissions from new and older design kerosene
heaters were investigated in a three phase research program.  This
presents  some of the results from Phase IT.  Phase II characterized
and particle emissions from four heaters representing the four types
tested in Phase I (1,2): the traditional convective (C) and radiant (R'
heaters as well as two dual-combustion chamber heaters, convective/radi3
(C/R) and radiant/radiant (R/R) .  The R/R heater had a catalyst.

                                  Methods

     Ten experiments, each lasting 13-1/2 h, were conducted in an
aluminum-lined 34 m3 environmental chamber.  One experiment, scheduled
between heater experiments, evaluated chamber air without a heater
present.  The C/R and R/R heaters were tested in triplicate, the R
in duplicate, and the C heater once.  For the nine kerosene heater
experiments, chamber temperature was kept below 30 °C by passing the
recirculated air over a cooling coil.  Coolant temperature was kept
the dew point of chamber air to prevent scavenging of combustion gases
condensing water vapor.   The air exchange rate, 1.210.1 air changes Pef
hour (ach) , was measured after each experiment by injecting CO  and
measuring the rate of clearance.  Complete mixing was maintained with a
recirculation rate of 95 ach.
     Heaters were initially burned dry then fueled outside the
with the same K-l kerosene used in Phase I.  At least 30 min before the
kerosene heater was lit, the heater was placed in the chamber on a Pott6
scale.  The heaters were lit and operated with a normal flame for 12 h  s
(1 h of buildup and 11 h of steady-state operation) .  Fuel consumption
measured in 100+0.5 g increments during buildup and every hour of
steady-state operation.

     Combustion gases (NO , S02, CO, C02> and HC) , air temperature, aa*1
dew point were continuously recorded on a multi-channel recorder
the 13-1/2 h of each experiment: 1/2-h prior to heater start-up, 12 h o
heater operation, and 1 h after heater shutdown.  Particle size        ,
measurements (electrical aerosol analyzer, optical particle counter, a°
condensation nuclei counter) were recorded for 10 min prior to heater
start-up, 30 min during heater start-up, 30 min every 3 h during       $s
steady-state, and 1 h after heater shutdown.  The rate of removal of Sa .
and particles by chamber surfaces, needed to calculate the emission ratgg
from chamber concentrations, was determined by comparing the decay of 8
and particle concentrations to that of C0? after heater shutdown.

     Samples of gaseous organic emissions were collected at 75 ml/min
hours 3 to 6 by a Bellows pump/SUMMA canister.  For the two heaters
in triplicate, vapors were also collected for the 1-h start-up and
periods (i.e., three 1-h start-up periods for each heater were
using one canister).  The organic compounds collected in the SUMMA

                                   770
-------
 Bisters were identified by  gas  chromatography/mass  spectrometry  (GC/MS)
at Radian Corp.

     Samples of semi-volatile organic  compounds  and particles were
         with a modified PM10  sampler (113  1/min) from  heater  start-up  to
        .  The PM10  sampler used  two tandem,  precleaned  [with
 *chloromethane (DCM))  filters  which were  changed after  6 h.  Both  filters
 *re  Teflon-coated glass fiber, but the second  filter collected particles
  •3  ym  more efficiently.  Semi-volatiles  were  retained  by  an XAD-2
 ^nister downstream  of  the filter.  Organic  compounds were  extracted  from
J16 filters into DCM by sonication; most of  the  extracts were used  for
 ado-mutagenicit.y testing at  the  EPA Health  Effects Research Laboratory
^pA-HERL).  Sorbent beds were  Soxhlet extracted with DCM and aliquots of
J16 extracts shipped to EPA-HERL  for Kado-mutagenicity testing  and  to
 riangle Laboratories for qualitative  GC/MS  analysis.

,    Samples of particles were  also collected by 24 portable Gilian pumps
   1/min) on four types of filters:

     1.   Eighteen pumps collected aerosol for  organic analysis on
         precleaned, Teflon-coated glass  fiber  filters  with a  Teflon
         backing.   The Teflon-coated  glass  fiber filters were  weighed for
         collected  particle  mass, the organic  compounds extracted into
         DCM by sonication,  and  the extracts composited.   Most of the
         extracts were used  for  Kado-mutagenicity testing  at EPA-HERL;
         not enough extract  remained  for  definitive  instrumental
         analysis.
     2.   Three pre-fired quartz  filters collected aerosol  for
         thermal-optical measurement  of organic/elemental  (o/e) carbon  at
         the Desert Research Institute and  at  Sunset  Laboratory.
     3.   Two PTFE filters collected aerosol  for elemental  analysis by
         x-ray fluorescence  at NBA, Inc.
     4.   Two phosphoric acid treated  quartz  filters  collected  aerosol for
         acid and inorganic  ion  analysis  at  Brookhaven  National
         Laboratory.   Only a subset of the  acid-treated quartz filters
         were analyzed for strong acid by Gran  titration,  ammonium by
         indophenol colorimetry, and  nitrate and sulfate by ion
         chromatography.

                                 Results

fl^  Fuel consumption (g/min) averaged 4.37  for  the C  heater [18,023 kJ/h
Ji7»lOO  Btu/h)], 2.50 for the C/R heater [9,908  kJ/h  (9,400 Btu/h)], 3.69
f°r the  R heater [13,175 kJ/h (12,500  Btu/h)], and 4.40  for the R/R heater
ji3»280  kJ/h (12,600 Btu/h)].   Emissions of CO,  C02, N0x> and S02 (yg/g
Opel) were typical of heater  design.   All  heaters produced  similar amounts
j.   c°2.  The R and R/R  heaters produced 3-10  times more  CO  than C and C/R
J"aters.  The C heater  produced 2-3 times  more N0x, predominantly as NO,
sjan  the C/R, R, and R/R heaters.  The R/R heater produced  about half the
  2 of the C, C/R, and  R heaters.

     Chamber particle concentrations,  as measured by  the portable pumps,
t  eraged 72, 105, 114,  and 1216 yg/m3  for  the C, C/R,  R, and R/R heaters,
,esPectively.  Background chamber aerosol  was 32±2 yg/m3 (1 standard
3!viation).  Particle emissions for the C, C/R,  R, and R/R  heaters were
 *• 42±4, 45±it and  376+47 yg/g fuel,  respectively.  The first  PM10 filter
v  llected 23±13% less mass than the portable  pump filters,  probably due  to
   atile stripping and  particle breakthrough.  Breakthrough was expected
     particles were less than  2.0 ym  in diameter with the  peak volume

                                  771
-------
 occurring  In  the  0,15  ym  size  range,  although  some  large particles were
 produced on heater  start-up.   The  second PM10  filter  collected  a visibl6
 deposit of 6-12%  of  the mass collected by  the  portable pumps.   The deg*e
 of breakthrough depended  on the heater: C  <  C/R  < R < R/R.

      Table I  summarizes the composition of kerosene heater aerosol.
 Inorganic  ion, carbon, and elemental  analyses  identified 99±14% of the
 aerosol mass.  Most, 45%  for the C/R  heater  and  54-60% for the  C, R.
 R/R heaters,  of the  aerosol collected by the portable pumps was
 principally in the  form of ammonium sulfate  (1).  Nitrate, at 0. 8-1-5
 yg/m3, comprised  0.1-1.4% of the aerosol.  Total carbon content, as     ^
 measured by thermal-optical analysis  at Sunset Laboratory, varied from
 to 47.2% of aerosol mass.  Aerosol carbon content for multiple  tests of
 the same heater varied from 4  to 14%.  Levels  of K, Ca, Ti, Cr, and F|  ,
 were  0.1-0.2  yg/m3;  aerosol from the  R heater  also  contained 3.2 yg/m
 Pb.

      Table II summarizes  the organic  emissions from the kerosene
 organic aerosol,  non-volatile  XAD-2 trapped  material, and gaseous
 non-methane hydrocarbons.  Aerosol color ranged  from black to yelloWi
 reflecting the carbon composition of  the aerosols.  Elemental carbon    f
 contents (reported as -C-) were 9.3%, 2.3-7.8%, 0.7-1.1%, and 0.2-0.3*  £
 the C, C/R, R, and the R/R heaters, respectively.   Organic carbon conte
 (reported as -CH -) was 35-38% for the C/R, 25% for the C, 20-22% for
                                                                       he
                                                                      ttl
R, and 3-8% for the R/R heaters.  Organic aerosol emission rates,       i
uncorrected for background aerosol, were 24±3, 17±2, 10, and 10 yg/g *u
for the R/R, C/R, C, and R heaters, respectively.  Organic compounds
R/R heater aerosol were 1.4, 2.0, and 1.2-2.0 times as likely to pyr
during thermal-optical analysis as organic aerosol produced by the C,  '
and C/R heaters, respectively.  The amount of material extracted from
heater aerosol, too small to be weighed accurately, was determined by
new liquid chromatograph/ laser light scattering method (LC/LLS) .
Sonication extracted 2.5±0.1%, 9.8%, 10.7+3.5%, and 17.0% of the R/Rt (jt
C/R, and R heater aerosols into DCM.  Estimated extraction efficienci6^
were 77%, 33-46%, 39%, and 22-37% for R, R/R, C, and C/R heater aerosol9'
respectively.

     The mass of material extracted from the XAD-2 canisters was
determined by LC/LLS and gravimetrlcally; the total chromatographabl6
semi-volatile organic material (TCO) was not determined.  The extract  ^
masses measured by the two methods were averaged since the LC/LLS me  .ie
is still under development.  These masses will be considered non-vol*
organic emissions.  Background chamber air contained 34 yg/m3 of
extractable material.  Chamber air during C, C/R, R, and R/R heater
operation contained 98, 24-100, 92, and 379-1192 yg/m3 of non-volati^-6 ^j
extractable material.  The effect of non-volatile material adhering c°
re-emitting from chamber walls on measured chamber concentrations couJ-
not be determined.  Estimated non-volatile organic emission rates were
11±2, 18±27, 18±5, and 74±62 yg/m3 fuel for the C, C/R, R, and R/R
heaters, respectively.  Extracts contained residual C8-C15 fuel
hydrocarbons, of similar or altered relative abundance, and combusti°£
products.  Data suggest that not all extractable material was detecta   ,
by GC/MS.  Non-volatiles emitted by the C heater contained more unalte
fuel and lower molecular weight combustion products- than the other     ^
heaters.  Non-volatiles emitted by the C/R, R, and R/R heaters cental*1
fuel with different amounts of lower molecular weight compounds remove
and different amounts of combustion products.  Fewer combustion  rodu
were observed 'if most of the 
-------
^Position of emissions from heaters operated more than once, the C/R and
 * heaters, were not reproducible.

    Volatile hydrocarbon emissions, as measured by the hydrocarbon
 °nitor during heater operation, were similar to background chamber levels
^d difficuic to determine accurately.  Volatile emissions were estimated
 0 be between 20 and 220 ug/g fuel for the heaters.  Interpretation of
t?st °f the SUMMA canister analyses was limited due to contamination from
 he analytical trap before the GC/MS.  The valid analyses identified
.""Pounds as components of unburned fuel and oxidized combustion products.
^Position was not reproducible but was qualitatively similar to the
Responding XAD-2 extract.  The highest concentrations, 2 and 7-8 times
 teady-state levels, were produced during heater start-up and shutdown,
 *8Pectively.  Gases collected during heater start-up and shutdown were
 aPorated fuel hydrocarbons, C8-C15 alkanes, and alkyl benzenes.

                               Discussion

e   Heater type and use affected the composition and amount of organic
/Batons more than CO , CO, NO , SO , or inorganic particle emissions.
f6r°sol carbon content varied from 5 to 47%; elemental carbon accounted
 r 3.5-30% of the carbon.  C and C/R heaters produced 2-9x more elemental
c,b°n than the R and R/R heaters.  The C, C/R, and R heater aerosols
b ntained 3 to 7 times more organic carbon than the R/R heater aerosol.
0°Vever, since particle emissions were 10-fold higher for the R/R heater,
J8atlic aerosol production was highest for this heater.  The abnormally
8l,8h Particle production and the poor reproducibility of R/R emissions
 *8&eat that the R/R heater was probably malfunctioning; the role of the
5n£alyst is uncertain.  Organic aerosol production for the C/R heater was
3 * higher than the C or R heaters.   Non-volatile emissions (yg/m3) varied
j* to 4-fold for the same heater and 11-fold between heaters.  Emission
C* (Ug extractable material/g fuel) was highest for the R/R heater.
cSJCal3-y. similar amounts of non-volatile and aerosol organics were
 aected.  Volatile emissions were difficult to quantitate.

„,.  Cifferences in the organic emission composition for different heaters
aJ3 replicates of the same heater were observed by thermal-optical carbon
 ^°s°l analysis and GC/MS analysis of XAD-2 and SUMMA samples, and
  8eeted by the different and variable extraction efficiencies of aerosol
^d non-volatile extractable masses.  The fluctuations between experiments
Kte reflected in both the non-volatile and organic aerosol emissions.
8i!,ConiP°sition of organic emissions of different heaters was sometimes
C Iar-  The differences within a heater were sometimes as large as
he Ve«n heater types.  Such variability precludes linking composition to
   er type and suggests that an unmeasured parameter is controlling the
   ustion process.

    Current data suggest that, for a given heater, lower aerosol
    lction correlates to higher organic content in aerosol, non-volatile,
   v°latile emissions that consist of unburned fuel and combustion
          The variability is not reflected in differences in fuel
          i.  The most obvious and non-measurable variable was the flame
        iu.  Data from Phase I indicate that particle and CO production
        without a substantial change in fuel consumption (1,2), but the
       data set is too limited to draw further conclusions.  Closer
          of current data and judicious re-analysis and extended analysis
   c°ntinuing.


                                  773
-------
                                References

1.   White, J.B., Leaderer, B.P., Boone, P.M., Hammond, S.K., Mumford»
     J.L. "Chamber studies characterizing organic emissions from keros^
     space heaters." In: Proc. of the 7th EPA/APCA Symp. on Measurement
     Toxic and Related Air Pollutants, Raleigh, NC, May 1987.

2.   White, J.B,, Leaderer, B.P., Boone, P.M., Hammond, S.K.
     "Characterization of particle and organic emissions from unvented
     kerosene space heaters."  In: Proc. of the 4th Intl. Conf. on
     Air Quality and Climate, Berlin, Germany, Vol. 1, 84-88, August
Convective
1
% of aerosol
SO, 60
OTT 21
nitrate 1.4
elements 0.8
carbon 36
% of carbon
organic 73
extract-
able 39
pyrol-
yzable 35
elemental 27
Convective/Radiant
123

39 40
94 92

37
24 25
6.2 8.0
Table II. Organic Emissions (yg organi
Convective
1
aerosol
o/e analysis 10
extractable 3.3
45
17
1.0
1.6
47
83

22
41
17
c/g
Convective/Radiant
123

18
4.3

16
2.9
Radiant
1 2
55
19
0.7
2.2
22 24
97 96

77
27 25
3.2 4.6
fuel)
Radiant
1 2

10
4.2
Radiant /Rao*- .
1 2
	 -"
58
24
o.i
7.2 5.8 5l?
98 95 9
£6
33
50 51 >[
2.6 4.8 3'
	 -"
-i — TtJf^
RadiantTKad1* 3
1 2
	 -*"

22
26 - 1
6.4
non-volatile
            a, b
            11±2
15±12   22±24
18±5  118150
volatile
   between 20 and 230
   Emissions not corrected for chamber background.
   Emissions not corrected for losses or re-emissions from chamber
   surfaces.
                                   774
-------
        FROM THE ENVIRONMENTAL RESPONSE
       EVALUATION OF THE TAGA 6000E DIRECT
    SAMPLING MASS SPECTROMETER / MASS
      H. Pritchett
  '• EPA Environmental Response  Team
  son, New Jersey

  ert E. Hague, Department  of Environmental
    Sciences, Rutgers University
V fT-
irj:; Winingham, formerly  of
  r* Research Consultants, Denver, Colorado

r^  The Environmental Response  Team evaluated the performance
  the TAGA 6000E Direct  Air Sampling Mass Spectrometer / Mass
  ~ rometer in order to  define  the various limitations of the
            This evaluation was undertaken to 1} address the
          various parties had raised concerning the
^  oiogy and 2) to define the  limitations of the instrument
de tJlat appropriate measures could be built into the ERT's
  6loping TAGA QA/QC plan.

V- .  Several limitations  were documented to include the
?ivation of instrument response both over the course of a
tj. 6h day and over a period of several days, the effect that
-------
Introduction

The history of the TAGA 6000E Direct Air Sampling Mass
Spectrometer / Mass Spectrometer has been mixed.  Some
reported successes in using the technology while others    1
stated "that operation of a mobile MS/MS system may be as *" e
of an art as a science" .  As will be shown below, without
proper Quality control and used in the improper applicati011
the technology is more art and salesmanship than science.

The instrument has several strong advantages which make  its
use very desirable if its limitations can be adequately
defined and compensated for.  The instrument has almost
instantaneous response for low ppb levels of many priority
pollutants - ideal for emission source identifications.  ^
uses a direct air sampling probe which avoids the various
problems associated with trapping efficiencies, desorptio11
efficiencies, and target compound losses and cross         j0ji
contamination during sample storage.  It has typical detect,0f
limits for priority pollutants in the low ppb range needed
environmental analyses.  All of these advantages make it   ^
worthwhile to find solutions to the real problems observed
many of the TAGA's critics.
                                                        „
The Environmental Response Team  (ERT) of the U.S. EPA has  e
owned its TAGA 6000E for approximately three years.  Over ^
past 18 months the ERT and its contractor have been
a serious of method evaluations on the instrument's
capabilities and limitations.  This data would then be u
better define the type of applications that the ERT would
its TAGA.  The data would also be used as a starting point
developing a workable TAGA Quality Assurance / Quality CO&-
plan.  The work focused just on the Low Pressure Chemical
lonization (LPCI) source because of the total lack of
sensitivity of the Atmospheric Pressure Chemical lonizati0
(APCI) source for the major priority pollutants.

Experimental

                                                          d >
The many of the various procedures used have been discusse
detail in the Summary of the TAGA Standard Operating and
Reporting Procedures for the Love Canal Full-Scale Air
Sampling study (Appendix A of reference 2).  Three of
procedures - instrument tuning, cylinder calibrations,    ^.
intermediate response factors - are briefly discussed bel°

                     Instrument Tuning

Each day the townsend discharge current is set to 10
and the source pressure is adjusted to .9 to 1.1 torr.
high voltage quadrupole power supplies are then allowed t0
warm up for at least thirty minutes.  A tetrachloroethyl61^,^
and trichloroethylene vapor mixture is then introduced to
sampled ambient air stream.  After the electron multiplief
voltage has been optimized, the 85+, 130+, and 166+ parent
                             776
-------
 thftS-are scanned for on each quadrupole.  During the scanning
 O~ ion intensities are optimized by adjustments to the rod
     ets while the quadrupole resolutions are set to yield
 rgntially equivalent mass peak widths at high height in the
    6 °f °*55 ~ °-85 AMU while maintaining reasonable mass
     shapes (i.e.,  no major splitting of peaks and semi-
        peak tops).   This is an iterative process since
        in the rod  off-sets will change the peak resolutions
 0^  shapes and changes is the peak widths will affect the
 Pa^erved i°n intensities.  Once an optimum set of quadrupole
 „ rameters have been found then a mass calibration is
          on the quadrupole.  Once both quadrupoles have been
      then the instrument is ready for compound calibrations
 ge -3 cylinders containing gas standard mixtures.   This
 tQ eral  tuning procedure is the "standard" tuning strategy
  c°nunended by the  manufacturer's senior scientists.3

                   Cylinder Calibrations

         calibrations are performed by performing serial
 _utions of gas standard mixtures (25 - 50 ppm of each
 .^Pound in dry nitrogen) into the sampled ambient air.    A
   *mum  of three non-zero and one-zero points  are used for
     calibration and from ten to thirty measurements  are made
        concentrations.   The concentration ranges  start from 6
          (depending upon the original concentration  range in
C- cylinder) and go up  to  at  least  50  -  100 ppb.  The
con  ntrations are  always performed  in  increasing order of
Sstntrati?ns witn the  zero point being  obtained first.  Two
  s of calibrations for each  compound  are performed each day
     at the start  of the day  and one at  the end of the day.

                 Spike  Recovery Analyses

     analyses are  performed as at least  two point serial
          (one zero and  one non-zero point concentration).  At
     thirty measurements are  made at each concentration and
   intensities averaged.  The average  intensities at at the
    point are then subtracted from the average intensities at
    non-zero point and  the resulting differences are divided
      appropriate  intermediate response  factors to yield ion
  £ spiked concentrations.  The ion pair concentrations for
    compound are then averaged and the measured compound
       concentrations are then divided by the actual spiked
   entration.

              Intermediate Response Factors

     response factors are derived from the two sets of
        factor data which were most recently acquired before
         the analysis for which they are to be used.   The
^  vations of the  intermediate response factors and their
^ ulting error bars, which are both covered in reference 2,
C0  Designed to yield a symmetrical error estimate of the
  Puted concentrations.   Therefore,  since the computed
                            777
-------
concentrations are obtained by dividing the observed ion   ^
signals by the appropriate response factors, the intermedia
response factors are derived by averaging the reciprocals °
the two bracketing response factors.  Again, a more
derivation and justification of this approach can be found
reference 2 which underwent an extensive external review
the various agencies associated with the Love Canal Emer
Area Habitability Study.  The equations used are as foll°tf
     IRF = 2 * (RF-L * RF2) /  (RF-j^ + RF2>
and
     %error = ± (RFn - RF7) /  (RF, + RF9)
                   _L     £>       J.     £*


Results and Discussion

Several limitations of the TAGA were documented during
months.  These limitations fit into three major categories1
varying instrument response,  interferences, and effects ot
non-standard tuning strategies.

                Varying Instrument Response

We found that the instrument  response varied throughout "k^1
day and through a given 2 week sampling effort.  For exam?1
on one day over fourteen hours the response factor for tnf
128/91 ion pair of chlorotoluene varied from a low of 20 *
counts per second (icps) to a high of 60 icps.  During ten
days of sampling at one site  the response factor for the
112/77 ion pair of chlorobenzene ranged over the frequency
distribution found in Figure  1.

These results were not unexpected based upon our past
experiences.  Typically the response factors increased
the beginning of the day to the end of the day.  Previous
had shown that the high voltage quadrupole power supplies
required at least thirty minutes of scanning each morning
before the mass peak resolution would become stable.      >
the power supplies for twenty to thirty minutes was found
be sufficient to cause the peak widths to decrease, thus
resulting in a general loss in sensitivity.  Evidently, a  -$
similar but slower warming up or hysteresis effect occurs s
where else in the overall ion path - either in the source
on the ion lenses themselves.  Although this change in
sensitivity through the day is correctable with the use of
intermediate response factors and their associated error
further work is warranted to  identify the sources of the
and then correct them.
The variation of response over the period of days was
expected.  The instrument uses ambient air as its chemica1 .^
ionization reagent gas and the formation of hydrated hydz"c $$
ions is a major, competing source of primary ionization °.fy
reagent gas.  The absolute composition and absolute humid1 *
of ambient air are not constant over a period of days.
                             778
-------
 Th
 jnerefore,  one  should not  expect the sensitivity of the
  strument  to remain constant from one day to another.

 j. s variation  of  instrument  response from day-to-day does
 a.Uiv trate  a common mistake made by many early users (and some
 . Qltors) of the TAGA,  because it illustrates the effect of
      matrix has  on  the  instrument's  response.   Whenever the
      matrix was  changed from the  ambient  air that  was  used
   the calibrations,  the  validity of those  calibrations  was
         jeopardized.   In the  past, TAGAs were calibrated in
       air and were  then used to analyze for compounds  in
   rices  such as zero air and  incinerator stack  gas  -  two
         with drastically different absolute humidities
   ative  to ambient air.   Additionally,  incomplete work  done
   ^S Advanced Analytics with their TAGA suggested  that
   stic changes in the C02 concentration could also  affect the
   trument's response.4

ijj s drift in instrument  response did  result in  the
(J?°fporation of beginning and end-of-day full calibrations
^v ^inrnm) and the  use  of  intermediate  response factors2  into
inj ERT's developing  QA/QC plan.  However, when  the
rJjermediate response  factors  are used to compensate for
f>6r nse  drifts, the  instrument  is capable of quantitative
   *~~~    e.   At  our  most recent activation, the percent
          measured over  three days for  eleven different
         ranged from  74%  to 174%.  The  reported spike
tv^°veries were only outside the acceptable range defined  by
  6 %error bars in only three  of the 57  measurements.

                       Interferences

   9f the major limitations of the LPCI source  is the
   titude of ionization modes  possible for many compounds.
     multiple modes result in the formation of  interfering
       which are  not  always expected.    For example, 1,4-
     e (MW=88)  will form  the parent ions at 87  (hydride
   action by N0+), 88+ (charge exchange), 89  (protonation),
   107+  (water cluster of 88  ).  The  formation of cluster
      ions can cause  a lower molecular weight alcohol to
j^rfere with higher molecular weight target compounds.   A
oj ticuiar bothersome  set  of clusters are the water clusters
QiJ:!16 M* and M+l+ ethanol parents which interfere with  1,1-
   "oroethane during  indoor air analyses.   Many compounds
    fragment in the source to yield parent ions capable of
   Ucing interferences.   For example,  ethyl ether, ethyl
   ate and isopropanol will all form a  fragment parent  ion at
   which forms  the 61+ water cluster.   This cluster ion
          with  the 61/27  ion pair for  1,1-dichloroethane.
         , the n+1 chlorinated alkanes  with chlorines on
        carbons will  ionize by the loss of HC1  to form the
         parent ion as the M+ ion of the n chlorinated
        Several halogenated compounds  will also fragment to
  ^m secondary parent ions which are identical  or of equal
  5 to the primary parent ions of other halogenated ions.
                            779
-------
Methylene chloride is interfered with by structurally
identical fragments of chloroform, 1 ,1 ,2-trichloroethane
trichloroethylene.  Its 85+ parent ion is also interfered v
by a parent ion of Freon-12 of equal mass.

These interferences make the accurate interpretation of eve
the simplest of mixtures a non-trivial matter.  In such a
case, almost every daughter ion spectra must be manually  f
deconvo luted based upon the operator's judgment of the otft
compounds which might be interfering - a judgment that is
based upon his interpretation of the total parent ion
Unfortunately, no database exists, except in the heads
senior, experienced operators, that lists all the expected
parent ions for each of the potentially interfering common
compounds .

These interferences can also greatly affect the quantitati
accuracy of the instrument as is illustrated in Table I.
1 ,1 ,2-trichloroethane and 1 ,1-dichloroethylene results     j
illustrate the uncorrectable {based upon strictly TAGA dat
interference of the two compounds for each other while tn^
vinyl chloride 63/27 data illustrates the correctable
interference by 1 ,1-dichloroethane.
                                                          ed
Because of these  interference problems, the ERT has devel°P
a stance that all of our qualitative identifications of
unknown compounds are based upon additional confirmatory
analyses of either time-weighted tenax/charcoal or Summa
canister grab samples.  In addition, the evolving TAGA
plan now calls for the periodic taking of confirmatory
canister grab samples.  In the future, we will be using an
board portable GC to periodically perform parallel
confirmation analyses.  By using confirmatory analyses
Summa canisters and/or portable GC analyses, it is
for many compounds (and in months will be for others) to
utilize the real-time capabilities of the TAGA to track tn
relative changes  in target compound concentrations over tl
(and/or location) and then periodically convert these
concentrations to absolute concentrations based upon the
results of the confirmatory analyses.

          Effect  of Non-Standard Tuning Strategies

We found that the "standard" instrument tuning procedures  ,
not always not result in standard quantitative and
performance of the instrument.  The quantitative
could  influenced by the optimization strategy of the
and the qualitative performance was influenced by instrum#  ^
parameters not normally adjusted during the  "standard" tun1

Two different TAGA operator's were used - each of whom ha o-
different strategy for optimizing the  instrument per format*  4
under  the same "standard" conditions.  One operator optimi*
the instrument for just sensitivity (as measured by the
magnitude of the response factors.  This tune was
                             780
-------
Characterized by a high  level of background noise which
gained over 100 icps.  The other operator first insured  that
P1© average background signals of non-present compounds
Remained below 20 and then optimized  sensitivity.  These two
J^nes were considered the "low noise" and  "high noise" tunes,
 esPectively.  The low noise tune resulted in more accurate
   more reliable quant itat ions as is illustrated in Table II.
    "spike" data used  in Table  II was  from the actual
ji^ibrations.  The  signals  for  each concentration were pulled
;r°to the calibration curve  and  were divided by the derived
response factors.   Consequently, this  data was not affected by
*nY  drifts in  instrument sensitivity.  The better accuracy
j^ing the low  noise tune is illustrated by the comparison of
:£e  uncorrected recoveries  for  vinyl chloride at 25 ppb  and
jpUorobenzene  at 6  ppb.  The more questionable quantitative
  lability of the  high noise tune is  best illustrated by
    aring the  effects  of the zero corrections on the
    robenzene  and vinyl chloride 25 ppb spikes.

    "standard" tuning  conditions, as they currently exist, do
    always guarantee reproducible spectra.  On six different
    during a  three week period daughter ion spectra were
    ired for several compounds  using "identical" instrument
Tuitions.  The  results for 1 ,1 ,1-trichloroethane , toluene,
f'4-dioxane, acetone,  and trichloroethylene have been included
  Table III to illustrate  typical results observed.  In
    ion, on one day the spectra were  repetitively acquired.
    spectra was normalized to  a base  ion which was the  most
  ense ion on the  first day.   This base ion was retained for
;n^t compound  throughout the time period even if the
^tensities of other ions became more  prominent.  The day-to-
    variation  in normalized intensities, which is shown  in
    e in, was considerable for all compounds except acetone;
    base peak  actually changed  in four out of five spectra.
     shifts in base peaks,  along with the high occurrences of
     al interferences, potentially make unreliable the  sole
JSe  of the currently available  computerized spectral searching
Algorithms.  The single day variations were much more
 6^sonable; no shifts  in the base peak were observed.

    affect of  the tuning strategy also illustrated a
    amental limitation of the current  TAGA methodology to
    ne whether a given background is "clean" or not.  Because
pf the need to maintain the same matrix for analyses and
^ibrations,  it is not possible to present the instrument
    a previously determined "clean" blank sample.  Currently
      is no universally  accepted measure of the background
 nstrument noise.   Therefore, when the  instrument  is reading
   ~  50 ppb of  acetone in  the upwind  analyses, the operator
  s no way of knowing beforehand that this sample  that he
   suring nis signal standard deviations on is indeed "clean"
   is contaminated.  If  the  background  is contaminated, then
   variation is  really  a  combination of the measurement error
   the variation of an  ever-changing sample and his reported
                             781
-------
detection limits will be biased high.  More than once we nafol-
been in the middle of background determinations (partially
detection limit determinations) and have seen major, sudden
shifts in the signals for selected target compounds.  The
issue of developing and validating means to determine the t
background noise of the instrument and to measure the
analytical measurement error in a non-clean background botn
warrant further study.

Conclusions

Several limitations of the TAGA were identified during this
study.  They included problems associated with changes in
instrument response, interferences, and the effects of non-
standardized instrument optimization strategies during the
"standard" instrument tuning.  The latter two types of      fi
problems can currently be addressed with the QA/QC strateg1
outlined above.  The ERT has developed a tuning optimizati°|V
strategy of stressing the maximum sensitivity for the mini111 f
amount of background signal.  This strategy results in 9rea a
quantitative reliability.  Further work is needed to devel°P
tuning approach which will yield reproducible daughter ion
spectra.

Because of changes in instrument response during the day/ a
least two sets of calibrations should be performed each da^'
These bracketing sets of response factor data should then b
used to both correct the potential biasing of the reported  
-------
oo
CO
c
IP

3
V

o
E
3
Z
                       30
                       24,
18


16


14




10


 8
                                FIGURE  1.   FREQUENCY DISTRIBUTION

                                           CHLGROBENZENE 112/77 RFs Over 10 Days
                                           /
                                            x
                                            x
                                               Xx
                                                 x
                               /
                              X
                                x
                                                    X X
                                                     x
                                         x
                                                        X
                                              x
//
       x
       x
                                                                        X
                                                 /
                                                    T
                                         T	!	1	1	1	r~"—r~^—r^	r
            4-0   50   60   70   80   90   100   110  120  130  14-0  150  160  170  1 SO
                                                  RF Range [N(i) - N(i-1)]
-------
                     .  RESULTS FROM COMBINED SPIKES OF CYLINDERS D i B
              (Demonstrates Quantitative Effects of Cross-Interferences)
 CYL;
I ID
:  D
i  D
i  D
;  D
.  B
,  B
I  B
:  8
'•  B

i SPIKE HftTID /
: COMPOUND
Soike Ratio 1:1 (DsB)

1,1-Dichloroethane
Carbon Tetracftloride
Ethyl flcetate
1, 1, 2-Dichloroet?iane
Benzene
tethylere Chloride
1,1-Dicflloroethylene
let r acn 1 oroet ^y 1 ene
Vinyl Cnloride AVERAGE)
V:nyi Chloride (62/27)
Vinyl Chloride (63/27)
Vinyl Chloride (64/27)
Spike Ratio 1:2 0>:B)

1, 1-Dichloroethane
Carbon Tetrachloride
Ethyl Acetate
1,1,2-Dichloroethane
Benzene
hethylene Chloride
i, 1-Dichiorciethyierje
Tetrachloroethylene
Vinyl Chloride (ftVERftSE)
Vinyl Chloride <62/27)
Viryi Chloride <£3/27>
Vinyl Chloride (64/27)
Spike Ratio 5:1 (D:B)

1, i-Dicnloroethane
Carbon Tetrachloride
Ethyl Acetate
1, 1,2-Dichloroethane
Benzene
ftethylew Cnloride
1, 1 -Dichi oroet hylene
Tetrach 1 oroet hy 1 ere
Vinyl Chloride 


11-DCE

112-TCfl Is)
112-TCfl (s)


12-DCfl

" '

VNCL (S)


11-DCE
CI Effect?
112-TCft
112-TCfl



12-DCfl

                                                784
-------
TABLE II.  EFFECT OF TUNE ON SPIKE RECOVERIES PERFORMED ON THE CALIBRATION DflTfi
  ITSEiF (i.e., Points of Calibration Were Divided By the Derived RFs)
1
I
LOW NOISE TONE i
*ZE30* SISMAL i
"ZERO" CQNC !
1
X RECOVERY (6 ppb) !
* SECCVE3Y (12 opb) !
% RECOVERY (£5 ppb) !
* RECOVERY (50 ppb) !
HISH NOISE !
"ZERO" SIGNAL !
"ZERO" COMC !
* RECOVERY (12 ppb) !
X SEC5VERY !£5 -jpb) '.
X RECOVERY (50 ppb) i
* RECOVERY (100 ppb) !
•Z£ROH CORRECTED ',
j
* RECOVERY  i
X RECOVERY (100 ppb) !
CBEN !

14.7 !
0.3 !
%.&* :
%.6X !
101. ox :
1
]

823.3 !
2.9 !
242. 3X !
£35.7* !
192.8* !
!
S
1
1
1
220.4* !
224. 8X !
187. 3X i
I
CTCL

15.0
0.4
32. 3X
9£.3X
100. 9X

323.5
2.0
179. 9X
179.9*
15&.4X
1&4.4X
172. IX
152. 6X
VNCL

£2.7
1.3
112.4*
108.4*
102.6*

1715
6.5
136.9*
114.2*
107.3*
110.7*
101.1*
100.7*
1,1-DCE

12.8
0.7
105.8*
102. 7X
101.4*

760
4.7
_,
124.5*
106.2*
105. 6X
104. 2X
98.1*
100.6*
! TCE

13.3
i.3
106. 9X
103. 1*
102.2*

479
6.8
120.1*
105.6*
105.3*
94.5*
92.8*
98. 9X
   CBEN - Chlorobenzene
   CTOL - Chlorotoluene
   VNCL - Vinyl  Chloride
   1,1-DCE -  1,1-Dichloroethylene
   TCE  - Trichloroethylene
                           785
-------
TABLE III.        VARIATION IN NORMALIZED SPECTRA FOR SELECTED TARGET
                 COMPOUNDS DURING ft SINGLE MY  I OVER SEVERAL DAYS
1
1
i
1
1
: COMPOUND
! lii-Trichlorcethane
•• lll-Trichloroetnane
! 111-Trichloroethane
' 1 1 1 -Tr i ch 1 oroet hane
i Toluene
! Toluene
i Toluene
! Toluene
! Toluene
:
! 1,4-Dioxane
! 1,4-Dioxane
: 1,4-Dicxane
! 1,4-Dioxane
! Acetone
! Acetone
! Acetone
! Acetone

' Trichloroethylene
• Trichloroetnylene
; Trichlorcethyiene
; Trichl oroet hylers
i Trichloreetr.ylene
;
ION
PAIR
99/99
99/S3
99/61
99/27
92/92
92/52
92/51
92/41
92/39

69/39
69/45
89/29
89/27
58/58
58/43
58/28
58/15

131/13
131/95
131/61
131/60
131/49
09/17

09/30
NORMALIZED
100.0
11.1
11.2
6.7
100.0
2.0
2.6
1.7
1.6
100.0
16,7
13.6
7.9
100.0
1.9
2.0
2.0
2.1
DATE
10/02
SUMMARY
10/08 10/09
ION INTENSITIES
DAY-TO-DAY
VARIATION
AVS
100.0 100.0 100.0 100. 0
46.8 59.1 78.1 42.4
40.4 49.6 95.8 42.2
21.3 31.2 46.9 i 22.8
AVERAGE *RSD FOR NON-BASE
100.0
4.5
6.0
4.5
4.5
100.0 100.0
16.0 21.9
12.4 17.7
6.9 14.6
11.6 17.7
! 100.0
i 9.3
i 8.1
i 5.9
! 7.5
AVERAGE *RSD FOR NON-BASE
192.8
100.0
0.0
0.0
100.0
35.7
5.3
5.1
196.0
100.0
16.3
9. a
100.0
43.2
5.1
5.4
63.6
100.0
18.2
40.0 34.1
100.0 100.0
10,5 15.5
1105.3
',100.0
! 12.1
15.2 15. B 13.8 ! 10.9
AVERAGE *RSD FOR NON-BASE
100.0
39.4
6.5
4.4
100.0 100.0
87.4 124.1
11.0 14.8
11.5 14.3
: loo.o
! 66.0
! 6.5
: B.i
AVERAGE *RSD FOR NON-BASE
100.0
23.6
8.4
3.8
22.8
100.0
50.3
23.8
17,7
38.1
100.0
80,0
60.0
0.0
100.0
100.0 100.0
150.0 107.1
102.8 114.3
63.3 0.0
227.8 64.3
1100.0
; 82.2
i 61.9
: 21.0
; 90.6
SD
25.4
30.7
15.1
PEAKS

8.2
6.0
4.7
6.2
PEAKS
73.4

6.6
5.8
PEAKS

34.6
3.8
4.0
PEAKS
44.0
41.8
31.9
73.4
* RSD
59.9*
72.8*
66.0*
66.3*
•
88.4* !
74.4* !
79.2* !
83.2* !
81.3*
69.7* !
i
I
54.2* !
53.5* '.
59.1*
1
1
52.4* !
44.6* !
48.7* !
48.6*
i
53.5* !
67.6* !
151.9* !
81.0* i
VARIATION OBSER^ '
OVER 6
AVG
100.0
40.9
37.9
19.5
100.0
7.2
10.5
6.6
8.6

56.2
100.0
16.0
10. 0
100.0
83.1
13.1
10.7

-
-
HOURS, 10'° ;
t
STD * *® \
3.5 a.*;
4.9 I*!!;
3.2 IB-2*,'
ia.6* ;
i
1.8 8*-* ,'
'•5 1A'2i
1.3 *;;
u I3'tf!
_f>!
7.2 i^i
,
5.4 33'*!
40 40. '* '
1.V -j i
SB.* ;
J
2.8 3.**;
l 7 13.2*
irf
0.3 2-j;
	 ^:
" 1
1
— t
1
— 1
** J
                             ftVERftSE 
-------
        OF PRODUCTS  FROM THE  PHOTOOXIDATION OF TOLUENE USING
     ANALYSIS
 ,
toQ '  Kleindienst,  P.B.  Shepson,  C.M.  Nero
^a llr°P  Services,  Inc.   Environmental  Sciences
  earch  Triangle  Park,  NC   27709
S.E
^'  &umdei and  D.V.  Kenny
Vi Adv&nced Analytics,  Inc.
  °ver,  MA  01810
u
V'  CuPitt
ll.g sPheric Sciences  Research  Laboratory
^s*  Environmental  Protection  Agency
  earch  Triangle Park,  NC   27711
   Toluene/NOx mixtures were irradiated   in a   22.7 m3 Teflon
   °kamber  operated in  a dynamic  mode.   The effluent  from
    fixture  was  passed  at   140  1/min  to  a   tandem   mass
         r configured  with an  atmospheric pressure chemical
          (APCI)  head.    Many  of  the  ring  fragmentation
   °ts  were  identified  and estimates of the concentratidns
  tPbtained using standards  which were   structurally similar
       observed  products.    The  sum  of  the yields of all
   °ts measured was 44%.    If  the  reported  yields  of the
      (16%)   are included,  a carbon balance of approximately
     °ktained.   The  product  data  indicate  that  the  ring
   ent»tion   products   are   generated   by  a  variety  of
                             787
-------
Introduction
     Toluene   is   one   of   the   most   important   and  pervasi t
reactive hydrocarbons   present  in   urban  air.    It   is  the
significant   single  aromatic   species  in   urban   air   arisl
primarily  from  automobile  exhaust  and  evaporative emissi°n ,
Ambient concentrations  in  the range 5-40  ppbv  can   be  measuf
regularly  in  urban locations [1J.
                                                              t
     The   detailed  mechanism   for  the  atmospheric  oxidatio"  .
toluene has continued to elude  atmospheric   scientists  alth°u j
substantial effort has  been expended  in this area.   The gen^r
nature of  the  atmospheric  oxidation  is  understood,  howe^6 '
The oxidation of toluene in the  troposphere is  initiated al"10 ' f
exclusively by  reaction with   hydroxyl radicals.    Reacti°n
toluene  with  NO2  and Oa  has  been  shown to be too
account for any significant  degree of  atmospheric
[2].    The   reaction   of  OH   with toluene can occur either  ^
hydrogen abstraction from  the methyl  group  or by addition ° hjfj
to the  aromatic ring.  Abstraction of  hydrogen from the  *1
group leads to  the  formation   of   benzaldehyde  as a
product with  benzyl nitrate  also  formed  but  to a much
extent.  Recent  evidence  indicates  that   the yield
abstraction pathway represents  8% of  the  total  reaction wit1
[3].
                                                            i t"
     Addition to the aromatic   ring has  been shown  to lea;L0i»
the formation  of cresols as well as  to products resulting *  g
the fragmentation of the aromatic ring  [4],    This leads
variety  of  oxygenated organic  products   having   up  to
carbon atoms.  Among the  products   in  highest yield   are
alpha-dicarbonyl   products,   methyl  glyoxal  and
Conjugated gamma-dicarbonyl products  (butenedial   and  4-
pentenal )  have  been   reported  as  well as a  variety  of °
fragmentation products  [5J.  However, yields of these
have not  been reported and the extent to  which these  pro
significantly improve   the carbon   balances previously
for the photooxidation  of toluene is  not  known.
                                                             to
     The  current   study  has  been  undertaken  in order ^e
quantitate  the  yields  of  minor  products generated  in ^
photooxidation  of  toluene.    For  this  work a   tandem ^ j0
spectrometer has  been  employed  in conjunction with a dyn&
flow  reactor.    The   reaction  conditions have been set
product  distribution   characteristic  of  a relatively
extent  of  reaction  (e.g.,  with  respect  to the maximum
concentration) to minimize the  extent  of  secondary
The carbon balance resulting under  these  conditions should ft
indicate to what extent other unknown processes are      '111


Experimental Methods

     A 5.1 ppmv toluene/0.94 ppmv NOX mixture was i
a  22.7  m3  Teflon  smog  chamber.   The  chamber was  surroun t
-------
     nce   time  for  this  chamber).     Reactant NO* was added
  •  nuously to the chamber through a dilution manifold  from a
  lr*der  of  NO  in  nitrogen.    Toluene was added by bubbling
I*, —acu  through  pure  toluene  at  273  K.    The  radiation
  efisity corresponded to a NOz  photolysis rate of 0.1 min- l .

i,a.  Most  major  organic  and inorganic species were monitored
^ ^8 standard instrumentation [6].   NO, total NO*,  and Os were
tQ,SUr>ed  using  continuous  monitors.   The inlet and effluent
i,  Uene  concentrations were measured  using gas chromatography.
 ^8Urement of  peroxyacetyl nitrate  employed GC with electron
        detection.     Aldehydes  (HCHO,  CHaCHO,  C6H5CHO) and
^TOa-dicarbonyl  species  (CHOCHO,   CH3COCHO) were  measured by
S6l.1Vatizing with 2, 4-dinitrophenylhydrazine  followed  by HPLC
tyl ration  and  quantitation.     Nitric  acid was collected on
ds °^ filters,  extracted with 0.01 mM HC1O4,  and quantitated
^ Citrate using ion chromatography.  Acetic acid was collected
5C bubbling  the effluent through water (pH 7) and quantitating
 etate  by 1C.

tjj   MS/MS analysis  of the effluent  mixture  was achieved using
J* TAGA 6000 triple quadrupole  mass spectrometer (Sciex, Inc.,
it, °nto,   Ontario,   Canada).    Operational  details  for  this
}6 ti'ument have been given previously [7].   Input for the MS/MS
t^ired a sample flow  of 140  1/min and  was transferred from
fy6 Reaction chamber using a 5-m length of 19-mm Teflon tubing.
J0  TAGA 6000 MS/MS  can produce  chemical ionization  (CI)  parent
%   spectra   or  collisionally   induced  dissociation  (CID)
^ Shter ion spectra.   This instrument was configured to employ
fa  &tmospheric  pressure  chemical   ionization  (APCI) head to
•  e^ate chemically  ionized species  for the individual effluent
             The use   of this   head is especially advantageous
    oxygenated  species,  since   these   are   detected  with
      ;ly  high  sensitivities.     The  sample  flow  from the
       passes  directly into the ionization region of  the mass
   -rometer   with    no   prior   collection,    concentration,
          or  other procedure  which could  degrade the sample
             To conduct a semiquantitative analysis of the CID
,lft.  —-v* products,  response factors from several structurally
V.a*ar   compounds   were  determined.  These  compounds included
  B~ial,  4_hexene-3-one,  and 2,4-hexadienal.    Response factors
    Determined  by  injecting a known quantity of the compound
    the APCI and determining the  peak area   of the  CI  parent
   -ts^ and  Discussion

   Table I   gives  a   partial  list   of  the  gas-phase compounds
       during  the  irradiation.     All   compounds   listed vere
       in   yields  of,   at  least,   one  percent.  This yield is
        a   reactive   loss   of  742  ppbv   toluene   during the
             In addition, MS/MS  analysis does  not discriminate
   between   possible   positional   isomers,  and  some  of the
   unds  listed in   Table   I  (e.g.,   4-oxo-2-pentenaL)  may
   sent more than  one   structural  isomer.   Benzaldehyde,  the
   abstraction  product detected,  was  present with a measured
    of 5%,   This value  is   consistent    with  the predicted
    (8%)   occurring by  abstraction L3], when considering the
        of  a small  yield  «0.1  %c/c)  of   benzyl  nitrate and
                             789
-------
 subsequent  reaction or photolysis of benzaldehyde .

                                                          •  n
      Addition of  OH to toluene can lead to cresol  formati0"
 to  toluene   ring-fragmentation products.   None  of the c
 were  measured in this study although their yield has previ°u .g
 been  reported at about 16% of reacted toluene [3].   Most of
 other  products   reported  in  Table  I are products from fi ^
 fragmentation reactions.   As seen in Table I  methyl gly0*6;.,,)
 the  organic  compound  observed  in greatest yield (by c& t0t
 followed  by  glyoxal and  4-oxo-2-pentenal .    The mechanis"1 gt
 the formation of these  products has  been given by ShepS011^
 al. [4] and involves the initial  attack  by  OH  on  the °r  *
 position  of  toluene,   followed  by  addition  of  O2  to f°* Op
 bicyclic  ring. The subsequent formation of an alkoxy radic^-1 ^
 this  bicyclic ring causes  the rapid decomposition of the  f^e
 to  form these products.   If the two oxygen bridge  ( f ormin£ ^
 bicyclic  ring)   occurs  between  the  3 and 6 positions °n 4-
 aromatic  ring, methyl glyoxal and butenedial (cis-2-butene^M^
 dial)   is     formed   (1:1   stoichiometry )   following   \j 4
 fragmentation.   If  the  bridge  occurs  between  the  1  * $>
 positions on the ring,  glyoxal and 4-oxo-2-pentenal are f°r rj./
 The  appearance   of  glyoxal  and  4-oxo-2-pentenal  in  ne tf
 equivalent   molar  quantities  suggests  that  this path ®*yctf
 significant.     Most  interest  has  focussed  on  the pr°d $t
 generated   from   the  adduct  formed  by OH attack  on the °r ^
 position  of  toluene,   since  addition  to  this  positi<>n
 reported  to  occur with  a  yield of approximately 80% [2].
                                                          ..^
      Sixteen  other  compounds  were identified (or tentat* ^,
 identified)  having estimated concentrations  of 1-8   ppbv (  '$
 0.8   %c/c).    Most  of   these  appear  to  orginate  from  ^
 fragmentation products  of  toluene  or  its   aromatic oxi^a „
 products  (e.g.,  o-cresol).   Approximately 25 ring fragment* »
 products were observed  in  this  study  which  accounted \fltf
 total yield  of about 40%  of the reacted carbon.   Numerous ° $
 products  (many   from  secondary  reactions)   were   undoub* ^
 present  and    undetected  due  to  their lower yields.   ^
 interesting  to note that  the yield  of butenedial  was  me»s  $$
 at  4  ppbv  (0.3   %c/c).    This is more than  a factor of 20  ^
 than  that  measured for   methyl glyoxal.    (Recall   that ^ ^
 mechanism  described above,   these  two  compounds  shotil* >gS
 present in roughly molar  quantities.)   This  disparity indic  Ot
 that  butenedial   is  removed  much more rapidly by reacti0  Ol
 photolysis than  is  methyl  glyoxal or that the net generati
 methyl  glyoxal   by  other   paths is significantly  greater
 butenedial .
     One objective of this study was  to set  reaction
to minimize to formation of secondary products.   However*  .
is a trade-off involved in this case.   On one   hand  suff*^'
conversion of  toluene is required to obtain a detectable. *
of products.   Conversely,  a  small  extent  of conversi0
required to  minimize the  extent of  secondary processes
formation of numerous hydroxy  compounds  in  fairly sign
yields tends  to indicate  that secondary  reactions  might
been significant in  the  reaction  system.    For
hydroxy-4-oxo-2-pentenal  (3-HOP)  was  observed with  »
yield  of  nearly  5%.    Formation   of   this    compound
conceivably result from a primary process in which the
group is retained following ring cleavage  or from  a
                             790
-------
                 a cresol.   Dumdei et al.  [5] have presented a
   iriism for the formation   of 3-hydroxy-6-oxo-2,4-heptadienal
 t^e  observed in 0.6% yield) which involves the addition of OH
 5f  ^e  para position  of toluene  followed by  cleavage of the
 tyft &tic  ring.     It  is currently uncertain whether a similar
 ||Qp °f mechanism is responsible for a large fraction of the 3-
  a°bserved.    The  relatively large  yield measured, however,
     to argue that this compound comes largely  from a primary
          Alternatively, the mechanistic simplicity of forming
     from  the  addition  of OH  to o-cresol  (analogous to the
         of  OH  to  toluene  to  form 4-oxo-2-pentenal) makes
      m of this  compound   by  secondary  reaction appealing.
   UH  + o-cresol  mechanism seems  feasible, however, only if
   Products do   not  undergo  significant  secondary reaction.
     also that  the observed yield of 3-HOP would represent 36%
 Pl>^ne assumed o-cresol  yield.)   Of course,  a combination of
 Of ??ry and secondary sources is also possible.   An examination
 H^t-   formation  rate  of  these  hydroxy  compounds  during a
 ^d    mode  irradiation of  toluene to determine if there is an
 
-------
[3]  R. Atkinson, W.P.L. Carter, A.M. Winer,  "Effects of
     Pressure of Product Yields in the NOX Photooxidat ion °f
     Selected Aromatic Hydrocarbons," J . Ph ys. C hem . , 87.'
     1605 (1983).

[4]  P.B. Shepson, E.O. Edney, E.K. Corse, "Ring Fragmentati0
     Radical Reactions on the Photooxidations of Toluene an
     o-Xylene," J.Phys. Chem.. 88.: 4122  (1984).

[5]  B.E. Dumdei and R.J. O'Brien, "Toluene Degradation
     Products in Simulated Atmospheric Conditions,"  N'ature.i
     311: 248 (1984).
[6]  P.B. Shepson, T.E. Kleindienst, E.O. Edney, G.R.
     J.H. Pittman, Jr., L.T. Cupitt, L.D. Claxton,  "The
     Mutagenic Activity of Irradiated Toluene/NO* /H2 0/Air
     Mixtures," Environ. Sci. Technol . ,  19: 249  (1985).

[7]    R.J.  O'Brien,  B.E.  Dumdei,  S.V.  Hummel,  R.A.
     "Determination  of  Atmospheric  Degradation   Products
     Toluene by Tandem  Mass  Spectrometry , "  Anal.  ChejLu'
     1329 (1984).

[8]   R.J. O'Brien,  P.J. Green,  N.-L. Nguyen,  R.A. Doty»
     B.E. Dumdei,  "Carbon  Balances  in  Simulated
     Reactions: Aromatic Hydrocarbons," Environ. Sci.
     17: 183 (1983).
Table I. Organic Species Measured during Toluene/NOx
         Irradiation with Yield Greater than  1%.
Effluent Concentration ( ppbv ) %
Species
CH3C(O)CHO
(CHO)2
4-oxo-2pentenal
Chromatographic
134
150
MS/MS
80
120
60
     C6H5CHO                 37

                                                           3>*
3-hydroxy-4-oxo-2-pentenal                  35

 peroxyacetyl nitrate        85

6-oxo-2,4-heptandienal                      20

     CH3COOH                 47             30             1§
                                                           1-°
      HCHO                   50                            *

   hexadienal                                9
                                                           1-°
hydroxy-oxo-hexadienal                       9
                             792
-------
      COMPARISON STUDY OF THE COMBUSTION ENGINEERING 8202A AND INTEGRATED
 M/IBIC  SAMPLE/PRECONCENTRATION   DIRECT  FLAME  IONIZATION  DETECTION   FOR
   °bient NMHC values from the  PDFID method ranged from 1.3 to 4.5 times higher than those
Vifi   contlntjous analyzer measurements.  The difference In measurement appeared to  be site
'Ity C.C' w'tn the difference increasing as the sample concentration decreased.  It was also apparent
\ tw    s ln the nydrocarbon mix of the sample affected the difference  in measurement between
%a ° ^hods.  This study points out the need  for further evaluation of continuous  hydrocarbon
^Q|ftL'rerTlents' as we" as the need to use caution  in utilizing data from GARB modified Combustion
         8202A analyzers due to their apparent underestimation of NMHC concentrations by
  e  Varyin9  amounts.    These  studies  may have  significant  implications  on future  State
\t1?.errtat|ori Plans, when ambient hydrocarbon measurements are used in air quality modeling for
    'nation of ln
       ion of long term ozone control strategies.
                                        793
-------
Introduction

   Measurement of ambient concentrations of non-methane hydrocarbon (NMHC) by local a11
regulatory agencies  is obtained from continuous analyzers employing chromatographic
and detection by flame ionization (FID). These analyzers, such as the Beckman 6800 and Como"^
Engineering 8202A  (8202A), measure total  hydrocarbons and methane, then obtain NMHC
subtraction. Measurements from these continuous instruments have been shown to be urirel
particularly at lower concentrations, due to a variety of instrument related problems. These Pr° ^
include the indirect, subtractive nature of the measurement process employed, non-unifon* ^
carbon response for different compounds due to oxygen interference, inadequate sensitivity'
interference from water vapor. CO
                                                                                     I*
      Recently, a new method for measurement of  total NMHC concentrations in ambler*   ,
Preconcentration Direct Flame  Ionization Detection (PDFID), was  developed at the Environ^  ..
Protection Agency (EPA) and at the Research Triangle Institute (RTI).  This method uses
trapping to pre-concentrate the NMHC while providing separation of the NMHC from the
The Quality Assurance division of the EPA has made a number of refinements to the basic sy *
which have resulted in improvements in system accuracy and precision.®  Recent studies
shown the PDFID system to have a uniform per carbon response to different hydrocarbon spec'6
to show good agreement when compared to analysis by G.C. "sum of the species". (1>
    During the summer of 1987, the Ventura County Air Pollution Control District (VCAPCD)
study to compare ambient NMHC measurements from their existing 8202A continuous analyzer*
PDFID analyses.
Experimental Method
   Side by side analyses of ambient air from a common sample manifold were performed at a
(El Rio) and an inland valley (Simi Valley) air monitoring site by a continuous analyzer (8202A)
PDFID technique.  Three hour integrated samples (6 A.M. - 9 A.M. P.S.T.) were collected in
Summa polished stainless steel canisters for later analysis by PDFID in  the VCAPCD lab
measurements were then compared to the real time 8202A measurements.

                                  PDFID Measurements

    Sample Collection.  The integrated grab samples were collected in 6 liter Summa
stainless  steel canisters.   The samplers  used  for collection  were  based  on  an EPA
recommended in REFERENCED  The samplers used in this study were modified slightly by the ^fl
of multiple valves and canisters and a pressure  transducer to  make  a  permanent record
pressure  of the canister during sample  collection.  The configuration of the sampling '
illustrated in FIGURE 1.

   gample Analyslp.  The PDFID system described In REFERENCE 1 was duplicated in the
lab for sample analysis.  The analytical system operation Is straightforward. In the sampling
sample is drawn through a cryogenic trap which is immersed in liquid argon (-186°C).  After <
the desired volume, a six port valve is switched to the inject  position, air and  methane are
flush out of the trap (about 20 seconds), after which the liquid argon is removed and trap
started.  The NMHC's are then thermally desorbed from the trap as it is heated to 90 degree9
flushed directly to the FID.  Equipment used is as follows: Shimadzu GC 9APF Gas
Shimadzu CR-5A integrator and Nutech Direct FID Sample Preconcentrator. Equipment
                                         794
-------
    operation strictly followed the  Information  presented !n the PDFID Technical Assistance
        (REFERENCE 1).
         Control/Quality Assurance.  Prior to the collection of samples, canisters are cleaned by
  ^ately flushing with humidified zero air and evacuating them.  Before final evacuation, humidified
     r from each canister is analyzed on the PDFID system.  Any canister showing a concentration
          0.020 ppmc* is re-cleaned and re-tested before use in the field. It was found that non-oil
      d piston vacuum pumps could be  used for canister evacuation,  rather than a large oil
      d pump.

 H/j,   CePtance testin9 °f the samplers was performed prior to use in the field. After cleaning, they
 t0  tested for contamination by sampling humidified zero air, then analyzed by the PDFID system.
 » De clean, analysis through each valve of  the sampler must show less than 0.020 ppmc NMHC.
      s were also tested for sample degradation by sampling a humidified propane standard.

   ° Gain further confidence in the precision of the samplers, two identical samplers were installed at
       location and four pairs of  collocated samples were collected.  Analysis of these samples
       an  average difference between  the  collocated measurements of 9,8% with  a standard
    'on of  e.4%.
  v   to use> the PDFID ana|ytical system was cleaned by passing humidified zero air until the
^o H S showed less than 0.020 ppmc NMHC.  Daily calibration of the system was accomplished by
JHJ! | uP'icate analyses of a 2.454 ppmc working propane standard.  Approximately once a month, a
  ''Point calibration with an NBS-traceable propane standard was performed to ensure linearity of
    stern and accuracy of the working standard.  Following calibrations and  analyses of high
    tration  samples, (especially those with a large percent of high  boiling  hydrocarbons), the
   ^ cleanliness was epsured by analyses of humidified zero air.

  . 'n order to assure the accuracy of the analytic system, Inter-laboratory comparisons  were
  0rrTled.  An Intertab comparison with the EPA's Quality Assurance Division of the Environmental
        System Laboratory consisted of twelve data pairs.  Four were humidified propane  audit
       and eight were ambient samples collected at VCAPCD's monitoring sites.  FIGURE 2 presents
           regression of the data pairs. An additional interlab comparison was performed on two
       mples between the VCAPCD's PDFID analysis and G.C. "sum of species" analysis by EPA's
         Sciences  Research  Laboratory.  This comparison  showed  good  results, with relative
     differences of 3.0 and 4.7 percent.

                            Combustion Engineering 8202A

     8202A NMHC analyzer uses gas chromatography and flame ionization detection  (FID) to
     quantitative real time analysis of ambient levels of methane (CH4) and total hydrocarbons
     NMHC is derived by the subtraction of CH4 from THC.
(w
   )
        analvzers utilized for this comparison  were modified by the California Air Resources Board
     Quality Assurance Staff.  The major affect of the modification was replacement of the "peak
     circuit with a true integrator.  These modifications were recommended by CARB to provide a
    accurate NMHC measurement by the 8202A.   Field operation of these analyzers  followed
    s Standard Operating Procedures. ®
         , concentrations of NMHC are reported In units of parts per million carbon (ppmc), which for a specific
   Uru3 Is the concentration by volume (ppmv) multiplied by the number of carbon atoms In the compound.
                                         795
-------
       Quality Control/Quality Assurance.    Much of the Q.C./Q.A. procedures followed
 recommended in the CARB  Q.A.  manual.*3)   Specifically, the following Q.C./Q.A. checks
 performed on the 8202A's:

    a)  daily inspection/weekly Q.C. check sheets
    b)  multi-point calibration with methane (6-mo interval)
    c)  automated daily zero and single point methane span
    d)  annual performance audit with methane (by CARB)

    Data from an 8202A was  not used for this study unless the analyzer appeared to be
 correctly and the daily span was within +/-15% of the expected value.
Results

   The comparison of measurements for the coastal (El Rio) and inland (Simi Valley) sites is Pr^s
in FIGURES 3 and 4, In all cases, the PDFID measured higher concentrations than the 8202A.  >n
Valley data show a very constant ratio between the two measurements, with the average PDnD
approximately 1.4 times the 8202A value. Close inspection of the Simi data show this 1.4 ra
comparisons made, except one, which show a PDFID/8202A ratio of 3.05.  The El Rio data
much more variability than the Simi Valley data.  The El Rio PDFID value varies from 1-53
times the 8202A value.
Conclusions
                                                                                    to
The initial analysis of Simi Valley data indicated that applying a correction factor of 1.4
8202A data would make the 8202A data compare favorably to the PDFID measurement.
close inspection of one Simi Valley sample shows a PDFID/8202A ratio far from 1.4 (3.05)-
further analysis this reveals an important finding.  It Is important to remember that In *  .
system, after the NMHC's are cryogenically trapped, the trap Is heated, causing the nvo
-------
Acknowledgments
^ I Would like to express appreciation to the County of Ventura and the Air Pollution Control District
  their support of this project, and specifically to Dr.  Harold Richter and Mr. Vinson Thompson of
       Mr. Dave Dayton and associates of Radian, Inc. for their technical support.
T
   erences
ii   McElr°y- V-L Thompson, H.Q. Richter, A Cryogenic Preconcentratlon - Direct FID fPDFID) Method for Measurement of
               . Research Triangle Institute, North Carolina (1985).

        m, R.K.M. Jayanty, F.R McElroy, V.L Thompson,
      Assistance Document for Assembly and Operation of the Suggested Preooncgntration Direct Rame lonlzatlon
        DFID) Unit. Research Triangle Institute, North Carolina (1986).
            Quality Assurance - Volume II - Standard Operating Procedures for Air Monitoring. California Air Resources
    ^rometric Data Division, (1978).
                SAMPLE IN
                           TO 8202A
                               FILTER
                               CRITICAL
                               ORIFICE
                            METAL
                           BELLOWS
                            PUMP
                                               ABSOLUTE
                                               PRESSURE
                                             TRANSDUCER
                                                           SOLENOID VALUES
                                                                               JCANI8TER
1 CANISTER
                            FKJURE 1: 8AMPLINQ SYSTEM FOR INTEGRATED SAMPLES

                                ORTHOGONflL  REGRESSION
                                                             ORTHOGONAL!  Y-flX»B
                                                                   H=—     1.0439
                                                                   r-
                                                                             .396799
                                 2       3
                             QflD/£Pf) PPMC
                                  FIGURE 2: EPA INTER-LAB COMPARISON
                                             797
-------
0.7
0«
U
0.4 -
0.3 •
O2 -
0.1 -
 0
    NMOC SIMI SEPTEMBER-OCTOBER 1987
               c.tran v> nro
                                             NMOC ELRIO JULY-OCTOBER
FIGURE 3: SIMI DATA
                                            FIGURE 4: EL RIO DATA
    CAM *05 RUI1 2
           ;:; 7 t ft * y I J
PKrtO
1
£
3
4
INT. OFF
TIME
0. 472
0.669
1 .045
9573
13756
93223
£4609
V
V
V
             TOTrtL
                                  1IIMIJ
                                          cone
                                             b.
                                             9.7443
                                            66. aj:JV
                                            17.4332
                      141164
                                           100
           CONCENTRATION »    a.7
             FIGURE 6:  SIMI VALLEY 09/12/87 PDFID/82021 RATIO = 3.05
      CflN  64 Sini 09/25/87

PKNO
1
2
3
4
5
INT. OFF
TIME
0.484
0. 561
0.639
0.376
0. 932

AREA
50710
30135
53402
3772
13391

NK IDMU

V
V
V
V

COMC
32.6295
19.3909
33.6491
2.4269
>.*036
                                                    NflME
              TOTAL     155410               Itttl

            CONCENTRATION  =    0.83<»PPnC;

             FIGURE 5: SIMI VALLEY 09/25/87 PDFID/82021  RATIO*
                                  798
-------
  ne Integrated Air Cancer Project: Overview and Boise Survey Results
 ]rry T.  Cupitt
            Sciences Research Laboratory
        R.  Fitz Simons
    ronmental  Monitoring Systems Laboratory

 esearch Triangle Park, NC 27711
 p    The Integrated Air Cancer Project (IACP)  is a long-term research
 f    Ct  wnicn  combines  and  integrates  the  resources  and  expertise  of  the
   r EPA research laboratories located at  Research Triangle Park,  NC,
    goals of the  research program  are  (1)  to  identify the  principal
 ty  c^nogens  in the air  to which humans are exposed,  (2)  to determine
      emiss^on  sources  are  the major contributors  to the atmospheric
*sti  n  of carc
-------
     During the  heating  season of 1986-87, a major sampling program was ^
carried out in Boise,  Idaho.  Boise was selected from a potential list °
more than 30 towns  and cities for several  reasons: (1) RWC was known to
be a significant contributor to the high aerosol particle loadings whic"1
normally occurred in Boise during the  fall and winter.  (2) There were ^
numerous sampling sites  available in the Boise area which seemed promi5"
ing for the objectives of this study.  (3) The terrain and meteorology
seemed appropriate  for extrapolation to other locations.  (4)  The lo°a
government and environmental agencies  expressed  strong support for the
project.

     Boise is the capital city of Idaho, with a  1980 population of
slightly more than  100,000 people.  The city is a center of state and
local government functions and is home to  a variety of corporate head-
quarters.  There are no large or heavy industrial sources.  The urbam2
area is located  along the Boise River, which flows through the city fr°
the southeast toward the northwest.  The valley  floor is approximately
2700 feet above  sea level.  The area is bordered on the north and east
mountains which rise to an elevation of more than 7000 feet.  To the
south and east, the land rises in a series of steps, called benches,
until  a broad plain  is reached at 450 feet above the valley floor.
Meteorologically, the wind flow during the sampling period should be
dominated by up-valley flow during the day and down-valley flow at ni9n

     The Boise field sampling program consisted  of two phases.  The
summer  phase was designed to provide an opportunity to obtain  a limited
number of samples with no RWC contribution. It also served as a test o'
the planned sampling scheme.  The experience proved invaluable and led
several modifications in the sampling equipment and procedures.  The
second phase of the  field sampling program was conducted from November
1986 through February 1987.

     The Boise field program during the heating season consisted of ^°\
ambient and residential sampling.  The ambient sampling was  conducted  a
three primary sites and  four auxiliary sites.  Sampling periods were ***
hours long, with changeover  times at 7 a.m. and 7 p.m.  There  were I3
sampling periods scheduled per week, and one period was dedicated to
calibration, maintenance, etc.  The residential  sampling involved a    fS
matched pair of homes each week.  Over the study, ten pairs of homes ^
sampled.  One of the homes in each pair used  a wood stove, a fireplace
insert, or a fireplace.  The other home did not burn wood.  Sampling **
conducted in 12-hour periods  identical  to those  at  the ambient sawpH"9
sites.  Sampling began each Saturday morning and terminated after the
nighttime sampling period which  ended at 7 a.m.  Wednesday.  For analy5'.
purposes, the eight  sampling periods were composited into four samp1eS'
week-end daytime, week-end nighttime, week-day daytime, and  week-day   ,
nighttime.   Whenever samples were collected at the residences, identic3
samples were also taken at two of the primary ambient  sampling sites.
The homes selected for the residential  sampling were matched for age»
size, etc.   All  of the houses sampled were non-smoking homes.
                                   800
-------
    The three primary ambient sampling sites were located in different
Jreas of the city.  One site, at Elm Grove Park, (EGP) was located in a
 esidential area and surrounded by homes which used wood as a heating
 °Urce.  A second primary site at a fire station (designated FS) was
       near a major intersection and was intended as a roadway site.  A
          primary site was located at an Federal Aviation Administration
/ansmitter site outside the populated area south of the city and was
 e$ignated RCAG.  The primary sites were thoroughly equipped to charac-
      the ambient air and to collect samples for bioassay and organic
"a lysis,  and  for  source  apportionment.  The bioassay samples  were
.,
  'lected with high-volume samplers, into which impactor stages had been
      to limit the collected particles to the range of 0 to 2.5 micro-
pr  -• s aerodynamic diameter.  These particul ate samples were collected to
I °vide the extracts for apportioning the mutagenicity, for use in
|. Sntifying potential carcinogens through bioassay directed fractiona-
s °ni and for compositing for use in a carcinogenesis bioassay.  The
5 Upce apportionment samples were collected with a battery of samplers.
 •"pies were taken for elemental  and ionic characterization of the
g r£sol particles.  Samples similar to those collected for bioassay were
^lflered for Carbon-14 analysis of the source contributions.  The ambient
4e£  Pollution was characterized for the fine (0 to 2.5 micrometers
j^ °dynamic diameter) and coarse (2.5 to 10 micrometers) aerosol  load-
pr*s.  Continuous monitors were used to measure the light scattering
etc  rt^es and tne Presence of ozone, carbon monoxide, nitrogen oxides,
^«  Meteorological measurements provided data on wind speed and di-
4rid    ' temperature, relative humidity, etc.  In  addition, aldehydes
Ofo  v?^atile organic compounds (VOCs) were measured.  Semi-volatile
sJrn1cs we""6 collected with XAD-2 in conjunction  with the residential
^Pling,

nnt  The residential sampling consisted of samples at ten pairs of
pstcned homes, together with identical  samples collected at the EGP and
f0rPr1mary sites.  During each sampling period, filter and XAD-2 samples
K  chemical  analysis and bioassay were collected  inside the home with
*PDl-'00d burnin9 appliance, at the top of the stack of the wood burning
"ut  !ance» inside the home without a wood burning  appliance, immediately
Sl^side the home without a wood burning appliance, and at the EGP and FS
%6S*   ^e interiors and neighborhood  of the residences were also
Sn^^terized through measurements of VOCs and aldehydes, aerosol  loadings
lj Composition, continuous monitors for carbon monoxide, nitrogen  oxides,
 9nt scattering, etc.

F41p There were four auxiliary sites situated across the community at the
^4m  °Unds (FAIR>«  Camel Back Park (CBp)«  Winstead Park (WINS),  and
dl£?s Elementary School  (ADAM).  Each auxiliary site was equipped with a
Cto"°tomous sampler  to  characterize the aerosol  mass  and elemental
            and with a high volume sampler to provide a sample for
        and  bioassay fingerprinting.   A special  study to  measure  and
        *ne  transport and mixing of the wood smoke plumes was also
  QUcted as  part of the IACP.

 r  ^n addition to the residential  sampling, a survey was used to learn
    about Boise's two major sources of PICs, namely  RWC and motor
                                  801
-------
vehicles.  Two survey  forms were used.  The  first  was  a general  survey
dealing with home heating and motor vehicle  usage.  A  second survey
administered to  respondents who burned wood, and dealt with RWC  in
detail.  Figure  1 shows the responses to questions about whether or no
the  residents owned  and used wood burning  appliances.  Sixty-two
of all homes in  Boise had a wood burning appliance.  Twenty-three
of all the respondents used their wood burning  appliance on a daily
during the heating season.  A total  of 40% of the homes had a firepla£e'
wood stove, or fireplace  insert which was  used  at  least twice per wee"'
Figure 2 presents data regarding the motor vehicle fleet in Boise as   t
described by the residential survey.  The  first cluster of bars  repre5
the  percentage of all vehicles for each of the three common types of
fuel, leaded gas, unleaded gas and diesel.   Almost 40% of all cars,  , .
trucks and motorcycles described by the respondents used leaded gasol'
and  about 3% used diesel  fuel.  The remaining clusters of bars repress
the  percentage of vehicle miles driven per day as a function of fuel  ^
type.  During the work week, the vehicle miles  profile is very s
the  number count data.  On weekends, however, the nunber of vehicle i"
for  vehicles using leaded gas increases.   This  change  is the result °
increased mileage by the vehicle described as "Car 2" or "Truck 2".   |6
Perhaps the "No. 2"  vehicle is more likely  to  be a recreational velllla,
or to be older.  The data from the survey  must be combined with info1"1"
tion  on the commercial vehicle fleet before  the total  picture of the
motor vehicle source for  Boise can be described.

      Sampling in Boise was completed in February 1987.  Analysis of t(1
samples and the data are still  underway.   Preliminary  results will  tie
reported in the other papers in this session.
                                   802
-------
Wood Burning Appliance Ownership & Usage
      Heating Season - Boise, Idaho
 Per Cent of Residences in Each Category

                 .38%
                          EZJ No Appliance Owned
                          a Appliance Not Used
                          M Occasionally Used
                          • Used 2-3 Times per Week
                          • Frequently Used
                          • Used Daily
    Figure 1. Representation of the fraction of homes in Boise, Idaho
    which own and use wood burning appliances, like wood stoves,
    fireplace inserts, or fireplaces.
 Non-Commercial Vehicle  Fleet
     Fuel Types and Usage
                                   I   I Unleaded Gas
                                   •1 Diesel
       Number of
       Vehicles
Vehicle Miles
 Mon - Fri
Vehicle Miles
 Sat & Sun
   Figure 2. Boise survey results for distribution of vehicles and
   vehicle miles as a function of fuel type.
                     803
-------
 INFLUENCE OF RESIDENTIAL WOOD COMBUSTION
 EMISSIONS ON INDOOR AIR QUALITY OF
 BOISE, IDAHO RESIDENCES
 V. Ross Highsraith & Charles E. Rodes, EMSL
 Roy B. Zweidinger, ASRL; Joellen Lewtas, HERL
 U. S. EPA, Research Triangle Park, NC 27711

 Anthony Wisbith, PEI Associates, Inc.
 Cincinnati, OH U52U6

 Richard J. Hardy, Morrison-Knudsen Engineers
 Boise, ID 83709

      A residential  monitoring  study  was   conducted in  Boise,  Idaho,
 November 1986  through  February  1987 to evaluate  the impact  of  re
 wood combustion (RWC) emissions  on  indoor air quality particulate, seIfllVQj'S,
 tile organic, and organic samples.   Samples were collected  indoors, outd ^
 and at the RWC  source  at 10 pairs of residences consisting of one ho&e  ^
 and one home  without an operating  RWC appliance.   Twelve-hour sampl*11^ ^ 2
 conducted at  each  set of  residences over  a  4-day period  (2  weekdays   et
 weekend days)  with indoor  sampling  occurring  in  the  room  where hoifle0 ^
 activity was highest.  Outdoor samples were collected immediately  outsit  ^
 residence without the RWC appliance.   Source  samples were  collected ff° pA$
 operating appliance  stack.   Key criteria  pollutants were  monitored
 indoor and outdoor  location.   Homeowner activity and  RWC appliance &c
 logs were  maintained.   The  analytical results  and homeowner activity
 have been examined to determine relationships  between key  pollutant conoe  3,
 tions, the operation of  the  woodburning appliance,  and homeowner  activ* ^e
 Fine particle  concentrations   inside the home  with the  RWC  appliance   s
 4-l6 ug/m3 higher than the  levels  observed in  the  homes without  applia tfj>
 Fine particle concentrations exceeding  100  fj,g/m3 were  observed in home3  ^i
improperly operated, misfueled, or  damaged  RWC appliances,  as  well as *•  $
 home with  an ultrasonic  humidifier  charged  with  municipally  supplie  >gd
 water.  The  sampling  protocol,  results,  and  conclusions   are  preSe
                                      804
-------
Production

of   Technological advances in  woodstove design and  the  relatively low  cost
area°°d hflve resulted  in increased  residential wood  combustion  (RWC)  in  many
«0ll s across the  United States.   Many  homeowners  are using  this  renewable
^BlCe 3S  3n alternative  to more costly fossil fuels.   Studies in.airsheds
Corrnated  by RWC emissions  show degradations  in ambient air quality directly
8tlJ.^sP0nding to the increased  use  of  woodburning appliances.1-lf  Laboratory
d|a  GS reveal   that  RWC emissions  are  rich  in fine  particle  (aerodynamic
t0lBl)eter <  2.5  urn)  polycyclic  organic matter  and  products  of  incomplete
irid0Ustion<      Several studies  have  evaluated the impact  of  woodburning on
*  ^or zifv* jvi*^k.'ij'.^__ H ^ in**      ^ * i       .«•                  m •  m   m.   - .
                                                     impact of woodburning  on
       3ir (luality«8~10  Many  of these studies were  resource  limited,  either
 t>r    nS on a  limited number  of chemical  species,  e.g. particles,  volatile
          etc., or conducted over short intervals.

     The Integrated Air  Cancer Project (IACP)  conducted an indoor air  pilot
      Curing 1985 in  residences  with  operating woodburning appliances.  The
      y objective of  this  pilot  study was to  develop  standardized  sampling
 lat analytical  methodologies  for future  indoor air  monitoring  programs.  A
 S0jg  Scale  residential study was subsequently planned for Boise, Idaho.  The
 itig   residential study was conducted in concert with an ambient air monitor-
 «m Pr°gram  designed  to  characterize  the  impact  of  RWC and mobile source
    sions  on the  Boise airshed.2

        oor» outdoor,  and source monitoring were  conducted during the winter
        season  at 10 sets of paired  residences, 1  residence with and 1  with-
 *^0i  Wooc*burning  appliance.   Only  homes without   smokers   were  monitored
 fesulnatinS  the high  concentrations  of  particulate  and gaseous  emissions
 Pos j**ng  from  tobacco smoking.  Paired residences were  matched  as  nearly as
 etc     with regards  to size,  construction,  building materials,  occupants,
 St^'   State-of-the-art particle,  volatile  and  semivolatile organic,  aldehyde,
 ffi8ir?  8eous  monitors  were  set  up and  operated continuously  inside  of  each
 ''Ma 6nce» immediately outside,  and  at  the flue of the RWC appliance,  over  a
   y  Period.
         ' Idaho, is  a  high altitude northwestern  metropolitan area  with  a
 *
-------
      Consecutive 12 h sampling periods  (changeover at  7  A.M.  and  7  P.M*) we
 conducted from 7 A.M. Saturday through  7 A.M. Wednesday.    Samples  for
 and inorganic  analysis  were  collected  on preweighed  Teflon   media     g
 standard PM^Q dichotomous samplers (total flow - 0.0167 m^/min).   Partid»la«,
 samples for bioassay analysis were collected on 102mm Pallflex  T60A20  Tef 1°" j
 impregnated glass fiber (TIGF) filters using PMjQ medium flow  samplers (O'|gj
 m3/min).  Vapor-phase  semivolatile  organic compounds  (SVOCs)  were collgC „
 with XAD-2 absorbent filled canisters installed immediately downstream °* ,„
 medium flow particle filter.   Cartridges  impregnated with 2,4-dinitrophe'^
 hydrazine and evacuated SUMMA polished  canisters  were used to collect **• .
 hydes and volatile organic compounds  (VOCs), respectively.  CO, NOX, and
 scattering (B8cat)  were  continuously  monitored at  each indoor  and  °u
 location.  Standard SF^ tracer release  sampling techniques11* were  used to *
 sure residential  air  exchange  rates.   Passive  nicotine  badges conta*n
 sodium bisulfate coated filters were placed in each residence  over the
 4-day period  to  monitor  for  cigarette  smoke. ib   Temperature and
 humidity were also monitored at each  indoor  location.   Particulate, SVOC,
 and aldehyde samples were collected at  the chimney outlets.  SVOC  and  bio
 samples were stored  at  -80°C  immediately following sampling.  Samples  for
 determination were conditioned for 24  h (20°C,  40% RH)  followed  by  st&
 gravimetric analysis.13  Aldehyde and  passive badge samples were capPe
twice the outdoor  concentrations.   The elevated potassium levels suggest ttf
contributions to the  indoor  environment.   The results of a  "best-fit"  Ai
regression between residential  fine particle mass and  fine particle  pot
yielded excellent  agreement  (correlation  coefficient  >0.92).   Indoor
lead levels  were  consistently  0.01-0.03  ug/m3   lower  than the  correspo0
outdoor lead  concentrations  with  no  differences  observed  between  P *
residences.  Pollutant  concentrations  measured  immediately outside the

                                     806
-------
  fen
  %e  V^kout  a woodburning appliance  were  equivalent to the ambient  levels
    rved at the two  fixed monitoring  locations.

  ljq_  Approximately J0%  of the  fine particles collected indoors were extract-
  fjsj ^th  dichloromethane  with  little  differences observed  between  paired
  "(to . Ces or  sampling periods.   Outdoor fine  particle  percent extractable
  26 J^?8 Averaged from  55% to 65%.  Indoor formaldehyde concentrations  (15 to
      '  were   U  to  5   times  higher  than  the corresponding  outdoor levels.
 &?*
 raf «- formaldehyde levels  in the  residence with the woodburning appliances
   ^re than  5  ppb higher than the average paired residence levels.  Total
   lQyla indoors (50-57 ppb C) were also U to 5 times the outdoor concentra-
      Indoor VOC concentrations  ranged between 500-1700 ppb C  and did not
 ,pe ^ to be  impacted by  woodburning  emissions.  Indoor  VOC concentrations
 .  •*• -5 to 2  times  higher than corresponding outdoor  levels.  Average outdoor
       mutagenicity (measured  as  the number  of revertants  [rev]  per  unit
      was nearly twice  the indoor particle mutagenic  activity.   The muta-
    activity  observed  in  the   residences  with operating  woodstoves  was
       elevated during weekend  daytime  periods,  again  corresponding to the
     °f highest  homeowner  use.   The  XAD-2  sample  bioassay results  were
     variable and  biased by a  large blank sample  mutagenic  contribution.

   CO concentrations at  the three  residential  locations for the seventh week
   nown in Figure  2.   The relationship observed between the 3 CO monitoring
  , *°ns (indoor with, indoor  without,  and  outdoor) is  representative  for
  ^  all the continuous parameters monitored  inside and outside at  the  20
    Residences.  In most cases,  the outdoor pollutant levels are  higher  than
         levels  with minimal differences  observed between paired  residence
       Increases in outdoor gaseous pollutant  concentrations were  generally
       by  reduced  indoor  concentration  maximums   that  occurred  after  a
n4tl _ time  delay.  Outdoor CO  levels were consistently 0.5 to 1.0 ppm higher
fyo-   levels indoors at the residence  with the woodburning appliance.   CO
    inside  the  woodburning residence were  0.5  ppm higher than the  paired
        CO levels.   Although not  significantly  different,  the increase  in
  Ir!centrations  in  the woodburning home is probably associated with  indoor
    8Ure 3  depicts  the  averaged  nighttime  fine particle  concentrations
    e and Q revealed
   st^^fferences between  indoor fine particle  concentrations.  A  review of
        data  for weeks  5  and 7  revealed improperly  operated  woodburning
         whicn allowed RWC emissions to escape into the indoor environment.
     fine particle  concentrations  in excess  of 100 ug/m3 were  observed at
        woodburning residences.  A comparison of the Bscat trace (Figure U)
       appliance log  maintained  by the homeowner  correlated  the  increased
       concentrations directly  to the charging  of  the woodburning appliance.
V 716 Particle  concentrations  observed  for  the week 8  residences  were
    ifficult to explain  initially.  The increased particle  concentrations
       at  the residence without a  woodburning  appliance were  totally unex-
        contradicted the original study design hypothesis.  After a compre-
   t   data review  and a return  visit to the  residences  in  question,  the
       fine  Particle concentrations were determined to result from the  use
   Uitrasonic humidifier charged with tap  water in a  sick child's  bedroom.
    the  Bscat  data were correlated  with the homeowner activity log,  which
          included  detailed times when the humidifier was turned on and off.


                                   807
-------
Immediate increases  in  particle concentrations  were observed when the "•  ^
owner turned  on  the humidifier.  Particle concentrations  gradually
after the  humidifier .was  turned off.   Indoor  fine  particle
were back-calculated  by  using  the  steady-state  equation  and  knowing  .
mineral content  of  the  Boise municipally supplied tap water, time  of  ^
fier operation,  volume  of house, air exchange rate,  and  humidifier
rate.  These calculations confirmed the observed fine particle concent ra
including the  nearly  100 (ig/m^ concentrations  observed during one  nig*1 .^
sampling period.  A. follow-up study has confirmed that  the ultrasonic W&- .t
fier was  the  source   responsible  for  this  increased  indoor  fine part
concentration.21

Conclusions

    The residential  monitoring  study  results  suggest  that  when
operated, woodburning  appliances do  not significantly  contribute part id ^
organic pollutants to  the homeowner's indoor air environment.   This  reia* ^
ship does  significantly change  when a  woodstove  is inisfueled,  damage^'  „
operated incorrectly.   Fine particle concentrations  exceeding 100 \i&/^ .$
be expected under these conditions.  Fine particle concentrations appr°aC ^
100 ug/m3 were observed when an ultrasonic humidifier  charged  with  taP_  gjt
was operated in  one residence.  Coarse  particle concentrations were "&•& ~
during daytime  sampling periods and are most  probably  associated with   ^
owner activity.   Indoor fine particle  concentrations  and  mutagenici'ty  j,
lower and indoor aldehyde and VOC concentrations were higher  than corresf  s
ing outdoor  levels.    Indoor  gaseous pollutant  concentration  maximum^   Q{
lower and slightly delayed  when compared to outdoor levels.   The result ^
this study  reemphasize the importance  of maintaining accurate  and  ^ ye
activity logs, as well as  fully documenting any emission sources that J"8*
in operation during the sampling program.

Acknowledgement
                                                                          0
     The authors  thank Dan  Fitzgerald, Greg  Meinors  and  Sonja  Feli*» ^
Associates, Inc.  for  coordinating the  residential monitoring effort ^
numerous personnel who contributed to this project.

Disclaimer

     The research described in this paper has been  reviewed by the
tal Monitoring  Systems Laboratory,  U.S.  EPA  and  approved  for publica
Approval does not signify that the contents necessarily reflect the
policies of the Agency, nor does mention of trade names  or commercial P
constitute endorsement or recommendation for use.
                                     808
-------
  *   V. R. Highsmith, C. E. Rodes, R. B. Zweidinger, R. C. Merrill, "The
     collection of neighborhood  air  samples  impacted by  residential wood
     combustion in Raleigh, NC and Albuquerque, NM," 1987 EPA/APCA Symposium
     .on Measurement of Toxic and Related Air Pollutants, Air Pollution Con-
     trol Association, Pittsburgh, PA, 1987, pp. 562-572.
 I
     V. R. Highsmith, R. B. Zweidinger, J. Lewtas, A. Wisbith, R. J. Hardy,
      Impact of  residential wood  combustion  and  automotive emissions  on
     the Boise, Idaho,  airshed," 1987 EPA/APCA Symposium on Measurement of
     .Toxic and Related Air Pollutants, Air  Pollution  Control  Association,
     Pittsburgh, PA,  1987, (In press).

     K. Sexton, J. D. Spengler, R. D. Treitman,  W. A.  Turner, "Effects of
     residential wood combustion  on indoor air  quality:   A case  study in
     Waterbury, Vermont, Atmos. Environ.  18; 1371-1383 (1984).


 *.   «
     R. K. Stevens, C. W.  Lewis,  T.  G. Dzubay, R. B. Baumgardner, L. T.
     Cupitt,  V. R. Highsmith, J. Lewtas, L. D. Claxton,  B.  Zak,  L.  Currie,
     "Source  apportionment of mutagenic activity of fine particles collected
     in Raleigh,  NC,  and  Albuquerque,  NM."  Presented  at  1987 EPA/APCA
     .Symposium on Measurement of  Toxic and Related Air Pollutants,  Air
     Pollution Control Association,  Pittsburgh,  PA,  1987.

     D.  G.  DeAngelis,  D.  S.  Ruffin,  R.  B.  Reznik,  "Preliminary character-
     ization  of emissions from wood-fired residential combustion equipment,"
     EPA-600/7-80-040.   U.   S.  Environmental  Protection  Agency,  Research
     Triangle  Park, NC.   Available as PB 80-182066  from National Technical
     Information Service,  Springfield, VA,  1980.

     B«  R.  Hubble, J.  R. Stetter, E.  Gebert, J.B.L.  Harkness, R.  D.  Flotard,
     'Experimental measurements  of emissions from residential woodburning
     stoves,"   Residential  Fuels; Environmental Impacts and  Solutions
     (edited by J. A.  CoopeY, D. Malek).Oregon Graduate Center,  Beaverton,
     °R, pp 79-138 (1982).
 7.
     s'  S. Butcher, E. M. Sorenson, "A study of wood stove particulate
     emissions,"  JAPCA  19;  724-728 (1979).
 8.
     D- J. Moschandreas, J.  Zabransky, H. E. Rector, "The effects of wood-
     burning on  the indoor  residential  air quality," Environ. Int. 4; 463-
     4&8 (1980).
 9,
     v- R. Highsmith,  R. B.  Zweidinger, R. G. Merrill, "Characterization of
     Indoor and  Outdoor  Air  Associated  with Residences Using Woodstoves in
     ^leigh, NC," Environ.  Int. (In press).
•0.
    *•  Sexton, J. D.  Spengler, R. D. Treitman,  "Effects of residential wood
    combustion on indoor  air quality,  A case  study in Waterbury,  VT,"
    Atmps. Environ.  18, 1371-1383, (1984).

 l'  »              ~~
    R*  G.  Merrill, D. B. Harris,  "Field and laboratory evaluation of a
    W0odstove dilution sampling system,"  1987  EPA/APCA Symposium on
    jkasurement of Toxic and Related Air  Pollutants, Air Pollution Control
    Association, Pittsburgh, PA,  1987.


                                  809
-------
12.  L. T. Cupitt, T.  Fitz-Simons,  "IACP  Boise field program:   Study
     and survey  results,"  1988  EPA/APCA  Symposium on Measurement of
     and Related Air Pollutants,  (In press) Air Pollution  Control Assod8*
     tion, Pittsburgh, PA,  1988.

13.  U.S.  Environmental  Protection Agency,  Inhalable Particulate
     Operations and Quality Assurance  Manual,  Environmental Monitoring   .
     Systems Laboratory,  U.S.  Environmental  Protection  Agency,  ResearC
     Triangle Park, NC (March  1983).

14.  R. L. Lagus, "Air leakage measurements by the timer dilution method* ^
     review," ASTM  Symposium on  Building Air Change  Rate  and  Infiltrat*°
     Measurements,  Technical Publication 719,  ASTM, (March 1978) p. 36***'
15.  S. K. Hammond, B. P. Leaderer, "A diffusion monitor to measure
     to passive smoking," Environ. Sci. Technology, No. 5, _2_^:494-497

16.  D. E. Lentzen, et al,  IERL-RTP  Procedures  Manual;   Level 1
     Assessment, 2nd ed. , EPA-600/7-78-201 .  Available as PB 193-795
     from National  Technical  Information  Service,  Springfield, VA
     pp. 140-142.

17.  B. N. Ames,  J. McCann, E.  Yamasaki,  "Method for detecting  carcinog6
     and mutagens with the  salmonella/mammalian-microsome mutagenicity
     test," Mutat.  Res. JJ1_: 347-364  (1975).
                                                                         if
18.  D. M. Maron, B. N.  Ames, "Revised methods  for  the  salmonella  mutag^11
     city test."  Mutat. Res. 113; 173-215  (1983).

19.  S. B. Tejada,  "Evaluation of  silica gel cartridges  coated  in-situ tf  e
     acidified, 2,4-dinitrophenylhydrazine  for  sampling  aldehydes  and ^e
     in air,"  Int. J. Environ. Anal. Chem. 26;  167 (1986).
                                                                        the
20.  F. D.  Stump,  D.  L. Dropkin,   "A  gas chromatographic  method f°r ..,,
     quantitative speciation of C2-C13 hydrocarbons in roadway vehicle c*
     sions," Anal.  Chem. 57; 2629-2634 (1985).

21.  V. R. Highsmith, C. E. Rodes, R. Hardy, "Influence  of portable hutnid
     fiers on indoor air quality," _E_S&_T_ (In press).
                                     810
-------
                                                      TABLE I     RESULTS Of INORGANIC ANALYSIS
CO
f SAMPLE
PERIOD
— -^— — — -^— ^
WEEKEND
DAY


WEEKDAY


WEEKEND
NIGHT


WEEKDAY
NIGHT

CONSTITUENT
FINE PARTICLES
COARSE PARTICLES
FINE K
FINE Pb
FINE PARTICLES
COARSE PARTICLES
FINE K
FINE Pb
FINE PARTICLES
COARSE PARTICLES
FINE K
FINE Pb
FINE PARTICLES
COARSE PARTICLES
FINE K
FINE Pb
ELM GROVE
PARK
25.4
6.5
0.09
0.04
27.1
9.4
0.10
0.05
43.4
5.9
0.15
0.04
36.4
7.1
0.16
0.04
FfRE STATO
— — — — __
21.2
9.5
0.06
0.05
— — ^ 	
26.4
22.8
0.06
0.10
36.6
9.4
0.12
0.07
37.9
13.3
0.13
0.07
•^•(^•i
IN
•MBH













rnn^m
RESIDENCE
WITH STOVE
— — — — ^—
14.8
0.20
0.03
20.6
9.9
0.15
0.03
27.5
7.1
0.13
0.02
24.3
5.3
0.14
0.03
^— - 	
RESIDENCE
WITHOUT STOVE
15.1
12.3
0.05
0.01
14.4
13.4
0.05
0.03
21.4
5.6
0.11
0.02
20.6
6.6
0.11
0.02
RESIDENCE
OUTDOORS ,
24.3
6.9
0.08
0.03
26.1
9.9
0.09
0.06
42.1
5.4
0.15
0.04
38.6
7.2
0.13
0.04
-------
                 Figure  1.   Residential monitoring sites,
            •   INSIDE WITH

            +   INSIDE WITHOUT
            0   OUTSIDE
       0700    1900    0700    1900     0700    1900     0700    1900
                               SAMPLE HOUR
Figure 2.  CO concentrations measurements  at the  seventh pair of resid
                                   812
-------
        AVERAGE OF FOUR  SAMPLING PERIODS
                                                          Wood burning
                                                          Non— Woodburning
                                                         Outside
         3.  Nighttime residential fine particle concentration (ug/m3).
>
 I

5


0

$
            INSIDE WITH



            INSIDE WITHOUT



            OUTSIDE
    °?00    1900     0700    1900    0700    1900    0700     1900


                           SAMPLE HOUR


     **•   Light scattering measurements  at  the  seventh  pair  of  residences.
                                 813
-------
DISTRIBUTION OF VOLATILE ORGANIC HYDROCARBONS AND ALDEHYDES
DURING THE IACP BOISE, IDAHO RESIDENTIAL STUDY
Roy Zweidinger, Silvestre Tejada and Ross Highsmith
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711

Hal Westburg
Washington State University
Pullman, WA  99164

Leslie Gage
Northrop Services, Inc.
Research Triangle Park, NC 27709
     The U.S. EPA's Integrated Air Cancer Project (IACP) conducted a
study in Boise, Idaho during November 1986-February 1987.  As part of
study, samples were collected in ten pairs of homes, with each pair
consisting of one with and one without a woodburning appliance.  Pa^
homes were located near each other and concurrent sampling was condu
inside each home and outside the home not burning wood.  A different
of homes was sampled each week during the study.  Averaged data front
homes showed total non-methane organic carbon (NMOC) and total carbo
be higher inside either home type relative to outside. Average forma
concentrations were slightly higher inside homes with woodstoves, but
several  non-woodburning homes had concentrations exceeding many of t"j
woodburning homes.  Benzene concentrations inside homes seemed relate0
mobile sources and were similar to outdoor levels.
                                   814
-------
Production

IVoi Between November 1986 and February 1987, the Integrated Air Cancer
thl«ec: » a  wood smoke  impacted site  located in  Boise's  Elm
  cont   i "* ^EGP) and a Dack9roiind  site  outside Boise at  an airport radio
  ind 7°   d a1r  to 9round  station  (RCAG).   Samples were  collected 7am-7pm
  *nd  pm"7am on  Saturday through Tuesday to  observe daytime vs.  nighttime
   u weekend vs.  weekday variations.

  Xpenmental  Methods

 AldBkHydr°carbons were collected in 6 liter  pacified canisters  (2).
 gepydes  were collected on dinitrophenyl-hydrazine  (DNPH) coated  silica
 US1 cartridges (3)  in duplicate.   All samples flow rates were maintained
   In9 mass  flow  controllers.

 fu  Hydrocarbon concentrations were determined by gas chromatography with
 ^r-B  i°nization detection.   Species in  the C2-Cc molecular weight range
 Oct* separated on a  packed capillary (20'  x l/16tf)  containing  Durapak n-
 f4n!ne/porasil C.  Analysis  of hydrocarbons in the C5-C10 molecular weight
 **     re Performed on a  30 meter DB-1  fused silica column.   Identities
     H
 an   determined  through retention time comparisons and mass spectral
           Hydrocarbon analyses in most cases  involved a composite of two
         samP]es representing the same time period of either a weekend or
        sa"ipling period;  eg.  Saturday and Sunday 7am-7pm were combined to
     °Je  weekend daytime  sample.   Typically, 0.5 liters of each sample to
       ned Were cry°9enically trapped on a modified  inlet system prior to
             derivatives of  carbonyl  species were eluted  from  the  sampling
          with 5 mL of acetonitrile and  analyzed by high  performance
      chromatography  (HPLC).  Two C-18 columns  (25 cm x 4.6 mm)  in  series
    Jfoployed using an acetonitrile/ water gradient and detection at 360
$tan,jA1dehyde identities were determined  by retention time comparison with
  "ndar
          I summarizes the average values for selected hydrocarbons and
         measured during the residential study. These averages represent
       we?ks of samPlin9 a"d individual values range over an order of
       e in some cases.  Median values in most cases, however, were within
(v/v?r the average values.  Total carbonyl s averaged between 50-60 ppb
'n
-------
 total.
      Formaldehyde concentrations  averaged 20 to 25 ppb  in  homes with wood
 stoves and 15 to 17 ppb  in homes  without stoves.  The highest  concentrat1"
 of formaldehyde observed  inside a home with a stove was  107 ppb while th*
 corresponding value for  a home without a stove was 33 ppb.  However,
 concentrations were not  always higher in the homes with  wood stoves.
 Figure 1 shows the concentration  of formaldehyde observed  during the    6
, weekend-nighttime sampling period for all ten pairs of homes.  Weeks 4,'iJ
 and 8 all exhibited higher concentrations in the home without  a wood $toV
 The week 8 "inrwith" home had a fireplace as opposed to  a  wood stove or
 fireplace insert which was the case in all other weeks.  This  home
 exhibited the lowest formaldehyde levels of any home in  the study which
 were nearly the same as ambient levels.  A fireplace normally  has a muc"  c
 higher draft rate than a wood stove and could result in  higher air excl"J|i
 rates for such a home.  However,  air exchange rates were determined f°r
 the homes in the study and were in the range of 0.4-0.7 hr"1.

      Figure 2 shows the formaldehyde concentrations for the sampling
 periods for week 7 of the study,  from Saturday morning through Tuesday
 night.   The "in:with" home that week was reported to have had a "leaky1
 wood stove as evidenced by wood smoke odor and elevated particulate
 loadings on dichot samples.   Due to a rather mild winter in Boise,  most
 wood burning activity occurred on the weekend.   This correlates with the
 observed increased formaldehyde levels for that  time period.   The home  f
 without a stove  evidenced more consistent formaldehyde concentrations °v
 the entire sampling period.   Outside concentrations were likewise fair^J
 constant with the exception  of Saturday and  Sunday mornings.   The increas
 levels  of Sunday morning  were also seen at the EGP primary site.
      Table I  contains the average total  non-methane organic carbon
 observed  for  the  residential  study.   Concentrations inside the homes
 about twice the outdoor levels with  notably higher levels being observe"
 weekend daytime samples.   The higher concentrations seen at the mobil6
 source site (FIRE)  during the weekday time  period may relate to increa$e
 traffic at that location.  Background levels were generally quite low- .
 Total  NMOC concentrations ranged  from 225 to 6721 ppb carbon (ppbC)
 homes  with stoves while concentrations in homes  without  stoves ranged
 476 to 9493 ppbC.   The  very high  concentrations   occasionally observed  js
 inside the homes were often due to elevated levels of one or two compoan
 For example,  the home exhibiting  total NMOC of 6721  ppbC had isobutane  ,
 concentrations exceeding  4000 and 2400 ppbC during the weekend and wee^
 daytime sampling periods  respectively.   Concentrations at nighttime  wef6
 order  of magnitude  lower.   Isobutane is  used as  a replacement for fr*°fl
 propel 1 ant  in some  aerosol  spray  cans  and use of such  may be an exp
 for these  high concentrations.

     Many  compounds associated  with  mobile  sources  frequently had s
 concentrations both inside  the  homes  and outside.   Ethylene   and be
 concentrations inside and outside the  homes  were  similar.   Isoprene
 concentrations, however, were much higher inside  either  home type than  5
outside and may reflect that  isoprene  is a  human  metabolite.   Isoprene %f
 also been  suggested as  a  potential tracer for tobacco  smoke(4),  but  ^
the homes  involved  in the study were occupied by  smokers.  Benzene
concentrations were just  slightly higher inside homes with stoves  than ^
those without.  The mobile  source site (FIRE) had elevated levels  re1»l
                                   816
-------
 show I! °ther ambient s!tes-  wnh the exception of the RCAG,  all  sites
 .... W6J a similar distribution pattern for benzene. Weekday daytime levels
     Highest followed by weekday nighttime.   Weekend nighttime
     "^ations were higher than weekend daytime concentrations.   The
        concentrations observed in Boise are similar to those seen in
       studies (1984-1986) of 39 U.S. urban  areas  (5).   Median values in
      studies ranged from 4.8 to 35 ppbC by  site with an overall  median of
     PpbC.

     The average toluene concentration inside either home type was about  60
     while  outside,  EGP and FIRE sites were  19-54  ppbC.   At the mobile
       site (FIRE),  the toluene to benzene concentration ratio held fairly
       * at about 2:1.   Inside the homes, however,  this  ratio increased to
        as  10 to one in some instances indicating  some  other  source of
        than mobile  sources.   More than half of the homes however,
        d the same 2:1  ratio seen  for the FIRE  site.

     usions

     1.  Concentrations  of  formaldehyde and other carbonyls were higher
    [e  homes  than  outside.
        Formaldehyde concentrations were higher  in  homes with wood  stoves
VnT  avera9e»  but not  in  all  cases.  Activities of  individuals,
CorJ! 1s"in9s»  etc.  likely are the major factors affecting carbonyl
  ncentrations in  homes.
°thp ''•.Ambient  benzene concentrations in Boise were similar to those of
PfJi  c*ties  studied.   Indoor benzene concentrations appeared related
  ^ominantly to mobile sources.
 Del
         research described in this paper has been reviewed by the
    ?Pheric Sciences Research Laboratory, US EPA and approved for
    1 cat ion.   Approval  does not signify that the contents  necessarily
 a 'ect the views and policies of the Agency nor does mention of trade
 se   or commercial  products constitute endorsement or recommendation for
       CuPitt,  T.  Fitz-Simons,  "IACP  Boise  field program: study design and
       results",  1988  EPA/APCA  Symposium on Measurement of Toxic and
   ated Air  Pollutants, RTP, NC,  1988.
2  I/
vjc; Oliver, J.  Pleil, W. McClenny,  "Sample integrity of trace level
can? ?le online  compounds in ambient air stored in "Summa" polished
  M1stersH, Atmos. Environ.. 1986, pp. 1403-1411.
3
     Jejada, "Evaluation of silica gel cartridges coated in situ with
      ed 2,4-dinitophenylhydrazine for sampling aldehydes and ketones in
      ntgrjL. jL Environ. Anal. Chem. . 26: 167 (1986).
4. G
0, ?• Lofroth, R. Burton, L.  Forehand, K.  Hammond,  R.  Seila,  R. Zweidinger
tohaW tas,  "Characterization of genotoxic components of environmental     '
       smoke",  Environ. Sci.  and Tech..  Submitted  May 1988.
                                  817
-------
5. R. Seila, "June-September, 6-9 AM ambient air benzene concentrations
39 U.S. cities", 1984-1986, Proceedings of the 1987 EPA/APCA Symposium'
Measurement of Toxic and Related Air Pollutants, RTP, NC, pp. 265-270.
                                                                        ifi
TABLE I.  Average Concentrations Observed for Selected VOC's
              During the Boise, Idaho IACP Residential Study3

WEEKDAY-DAYTIME
Formaldehyde
Ethyl ene
Isoprene
Benzene
Toluene
Total NMOC
Total carbonyl
WEEKDAY-NIGHTTIME
Formaldehyde
Ethyl ene
Isoprene
Benzene
Toluene
Total NMOC
Total carbonyl
WEEKEND-DAYTIME
Formaldehyde
Ethyl ene
Isoprene
Benzene
Toluene
Total NMOC
Total carbonyl
WEEKEND-NIGHTTIME
Formaldehyde
Ethyl ene
Isoprene
Benzene
Toluene
Total NMOC
Total carbonyl
INrWITH

24.3
34.0
7.0
24.4
68.9
1378.6
52.6

20.7
31.1
7.2
22.3
57.2
1009.1
51.1

25.6
28.8
4.8
18.3
57.7
1692.8
57.4

22.4
29.3
10.6
19.2
60.1
936.1
52.9
IN:W/0

15.4
30.4
6.7
19.7
71.0
1002.6
57.5

15.2
28.9
6.6
16.0
63.4
878.2
50.2

17.1
20.9
6.4
13.6
64.6
1737.1
51.9

16.5
26.4
8.5
15.8
66.3
1049.8
52.4
OUTSIDE

4.2
26.8
1.3
17.3
37.6
523.7
10.6

4.9
26.8
1.3
16.3
32.4
494.7
12.4

3.8
19.7
1.0
11.8
23.2
370.0
10.1

4.5
24.4
0.8
14.8
27.5
438.6
11.6
EGP

3.7
44.9
1.0
14.6
29.2
480.2
9.2

4.4
52.4
1.6
14.0
25.9
461.7
11.0

3.5
33.3
0.8
10.5
19.4
389.5
14.4

4.7
47.8
1.2
13.5
24.6
450.6
13.4
FIRE

4.8
37.8
2.6
22.5
54.1
803.2
11.2

4.7
40.5
2.3
18.2
42.2
688.3
15.6

3.3
23.4
1.9
13.1
28.1
447.1
17.3

4.4
36.5
1.9
15.8
31.9
443.1
21.5
RCAG

0.9
6.4
o.o
2.1
2.9
132.5
4.1

1.1
6.5
0.0
2.2
3.0
112.4
4.5

0.9
7.0
o.o
2.4
3.1
94.8
3.8

0.7
5.9
j*
o.o
/*
2.0
2.3
87.1
4.4
	 — '
a. Formaldehyde and total carbonyl concentrations are in ppb (v/v)
   Hydrocarbon and total NMOC concentrations are in ppbC.
                                   818
-------
Table II.  Average Percent of Total  Carbonyls
           Boise, Idaho IACP Residential  Study

    CARBONYLIN:WITH  IN:W/0OUTSIDE
 Formaldehyde        41.8     31.9     39.2
 Acetone             22.7     26.6     18.4
 Acetaldehyde        18.2     20.6     20.7
 Hexanaldehyde        2.8      2.8      1.0
 Propionaldehyde      2.6      3.1      3.4
 Butyraldehyde        2.1      6.2      3.8
 Valeraldehyde        1.5      1.4      0.6
 Benzaldehyde         1.3      1.2      2.0
 Acrolein             1.2      1.1      1.7
 Crotonaldehyde        0.4      0.2      0.5
 p-Tolualdehyde        0.4      0.2     0.2
 I sovaleraldehyde     0.3      0.4     0.1
 m-Tolualdehyde        0.3      0.1      0.3
 o-Tolualdehyde        0.2      0.0     0.0
 Unknowns             4.2      4.2     8.4
                    819
-------
                                          IN: WITH

                                      EZ2  IN: W/0

                                          OUTSIDE
                       456
                       SAMPLING WEEK
 Figure 1.  Average   weekend nighttime formaldehyde
       concentrations  during boise  residential study
                                          mm IN: WITH

                                          K23 IN: W/0

                                          eza
       AM    PM
         Sat
AM    PM
  Sun
AM    PM
  Mon
AM
   Tue
Figure 2.  Formaldehyde concentrations for each

          period of  week 7; Boise residential  study
                          820
-------
 emivolatile and Condensible Extractable Organic Materials Distribution in
        Air and Woodstove Emissions
    - Merrill Jr.
    an Corporation
  search Triangle Park, NC 27709
b ^
».* B. Zweidinger and J. A. Dorsey
K • S. Environmental Protection Agency
^search Triangle Park,  27711

J-F- Martz
Jciirex Corporation
^search Triangle Park, NC 27709
T v
o1^- Koinis
. °uthwestern Laboratories
q°Uston, TX  77054
   Information is provided in this paper on the distribution of vapor phase
s. Condensible extractable woodstove and mobile source emissions. The
  dies provide information from the field acquisition and laboratory
i     S °f samPles taken simultaneously from sources inside residences and
     surrounding ambient air.  Samples were acquired with techniques
  !gned to provide the same distribution of vapor phase and condensible
       at each site. Observations on the relationship of vapor phase
   Volatile and condensed organic material in the samples arc made. The
     ance of the semivolatile materials in assessing source impacts and air
     will also be discussed.
                                821
-------
Introduction

   A major sampling and analysis effort was undertaken by EPA in Boise,
Idaho, to study the effects of woodstove and mobile source emissions on air
quality. The sampling effort was conducted from November 1986 to
February 1987 and involved collection of airborne materials from:  inside
residences with and without woodstoves; woodstoves associated with the
residences; and ambient air immediately outside these residences, in a
residential park, and near a major traffic artery. Samples were collected
with 0.11 m^/min (4 cfm) sampling systems employing Pallflex Teflon
coated fiberglass filters followed by Amberlite XAD-2 sorbent resin. The
system used for sampling the woodstove emissions also employed a chamber
prior to the filter which diluted the sample approximately 20:1 and reduced
its temperature to the prevailing ambient level (1).

Experimental

   The Teflon coated filters were used "as received."  After each 12 hour
period they were removed from the sampler and stored in a cryogenic
freezer prior to shipment.  The XAD-2 resin was cleaned, loaded into sealed
sampling cartridges containing approximately 200 g each, and  shipped to
Boise in heat-sealed Teflon bags. The bags were opened at the  sampling site
and the cartridge installed in the sampling system. After each 12 hour
sampling period, the cartridges were capped, resealed in Teflon bags, and
stored in a cryogenic freezer until shipment. All samples were returned to
the EPA laboratory at Research Triangle Park under dry ice and extracted
within 4 days of receipt.

   Filter samples were recovered with a 24 hour Soxhlet extraction using
dichloromethane as the extraction medium. XAD-2  was extracted while stiu
in the sampling cartridge using dichloromethane followed by methanol in
continuous flow elution technique developed for the IACP program.
dichloromethane and methanol extracts were kept separate for  analysis,
the filter and XAD-2 samples were analyzed separately. Sample extracts
were stored at -80°C prior to analysis.

   Within a group of filters  or XAD-2 samples, pairing was performed to
composite extracts which represented different sampling periods. Each v^6
there were 4 sampling days - 2 weekend sampling days and 2 weekday      -
sampling days. For each day a "Day" (7 am to 7 pm) and a "Night" (7 pmto
am) sample was taken at each sampling site. The sample pairing was:
                                 822
-------
       Saturday & Sunday, 7 am to 7 pm - Weekend Day
       Saturday & Sunday, 7 pm to 7 am - Weekend Night
       Monday & Tuesday, 7 am to 7 pm - Weekday Day
       Monday & Tuesday, 7 pm to 7 am - Weekday Night

   Combined samples were concentrated using the Kuderna Danish method
     a three-ball Snyder column. XAD-2 samples were filtered to remove
     3 resin particles using 0.45 um syringe filters and a 10 ml Luer-Lock
 fringe. The paniculate samples were diluted to 5.0 ml and XAD-2 samples
 010.0 ml.  Samples were then transferred to Teflon lined screw-cap vials
 and stored at -8QOC until analyzed.

   An aliquot of each paniculate sample was dried to constant weight and the
 Gravimetric extractable organic matter (GRAY) determined. An aliquot of
 a°h  XAD-2 sample was also analyzed for non-volatile organic residue by
 jjravimetric procedures (2). XAD-2 aliquots were further analyzed to
 eterrnine total chromatographable organics (TCO) as a measure of the
 .^volatile content (2). TCO were determined using  a GC/FID calibrated
 °r hydrocarbons.  Results were calculated in terms of micrograms
 arbon/cubic meter of gas sampled. The results referred to as "Gravimetric"
 hjp the sum  of the condensibles from the paniculate and XAD-2.
 ^mivolatiles" are the TCO values from the XAD-2.

 Results

 e  Average "Day" and "Night" results for the Weekend and Weekday sets at
 Si^h of the five  sites are shown  in Table 1.  Individual data points for typical
   are presented in Figures 1 through 3.

   The ambient sites (EGP, Fire, Out) had the lowest average  concentrations
   "*~1  extractable organic matter (EOM), ranging from 52.6 to 122.8
       Concentrations at the Fire site were normally higher than at EGP or
^ut sites.  The highest concentrations of EOM were generally found on
 eekdays during the day. Further, concentrations of semivolatiles  were
 ^ays higher at these sites during the day on weekdays than on weekends,
 Resting an increased contribution from mobile sources during these
 nods. Semivolatile concentrations exhibited a wider range (22.7 - 84.4
 §/m3) man gravimetric concentrations (27.5 - 63.7 ug/m3).

   There was also a consistent pattern in the weekend samples which showed
5 rrcent of gravimetric materials higher than semivolatiles during the day and
 eversal at night. This same pattern was true for the EGP weekday
                                823
-------
samples.  However, the pattern was not seen for Fire and Out weekday
samples.  The semivolatiles were the major species at the Fire site for both
the day and night weekday samples.  This again is probably the result of
semivolatile emissions from mobile sources impacting on the daytime
sample.

   Concentrations of EOM inside the residences were two to three times
higher than at the ambient sites, ranging from 209.5 to 328.1 ug/nA
Weekend concentrations were always higher than weekday concentrations,
probably as the result of greater activity in the home on weekends. There did
not appear to be a significant difference in EOM between homes with
woodstoves and those without.

   The percent of semivolatiles in the residential samples was much greater
than that  of the ambient sample. Semivolatiles in the residential samples
averaged 78% as contrasted with 52% in the ambient samples.  Residences
without woodstoves seemed to have somewhat higher semivolatiles than those
with woodstoves, averaging 239 ug/m3 vs 195 ug/m3 for those with.

   Stack sample concentrations were between 3 and 4 orders of magnitude
greater than the ambient and residential concentrations. Daytime sample
concentrations were greater than nighttime by 95% on weekends and 88% °n
weekdays. The split between semivolatile and gravimetric species was a very
consistent 17% to 83% and did not exhibit any weekend/weekday or
day/night variability.

   The importance of including the XAD-2 resin in the sampling system is
illustrated in Table 2. For the ambient and  indoor sites the majority of the
sample (76% and 94%, respectively) was collected by the XAD-2. Even for
the stack  samples, which contained more EOM on the paniculate than the
XAD-2,41% was collected by the XAD-2.  The importance of these sorbent
collected materials with respect to their contribution to mutagenicity has not
been determined at this time. However, preliminary data indicate that the
organics collected on the XAD-2 produce a significant response in bioassay
test.
                                 824
-------
 Conclusions

   o Weekend samples taken at the ambient sites (EGP, Fire, Out) showed a
 ^finite day/night pattern of semivolatiles/particulates.

   o Semivolatile concentrations at the ambient sites had a much wider
 Variation than gravimetric concentrations.

   o Samples taken at the urban site (Fire) on weekdays indicated an impact
 °* mobile source semivolatiles.

   o Indoor concentrations of extractable organic matter were dominated by
 ^ semivolatiles.

   o Extractable organic matter was approximately equal in residences with
 ^ without woodstoves.

   o The XAD-2 collected organics dominated the ambient and indoor
 sainples and were a significant percentage of the stack samples.

 Terences

 •  V. Ross Highsmith, Roy B. Zweidinger, Raymond G.  Merrill,
   "Characterization of Indoor and Outdoor Air Associated with
   Residences using Woodstoves: A Pilot Study." Environment
   International, in press.

2l  E>. E. Lentzen, D. E. Wagoner, E. D. Estes, W. F. Gutknecht, "IERL-
   RTP Procedures Manual: Level 1 Environmental Assessment (Second
   Edition), EPA-600/7-78-201, PB 293-795, October 1978.
                                825
-------
                Table 1.  Concentrations of Semivolatile
           and Gravimetric Mass in Micrograms/Cubic Meter

                                 WEEKEND       WEEKEND
                               DAY  NIGHT   DAY  NIGHT
EGP
Fire
Out
In Without
Woodstove
In With
Woodstove
Stack
    Semivolatile
    Gravimetric
    Total EOM

    Semivolatile
    Gravimetric
    Total EOM

    Semivolatile
    Gravimetric
    Total EOM

    Semivolatile
    Gravimetric
    Total EOM

    Semivolatile
    Gravimetric
    Total EOM
22.7
29.8
52.6
37.3
44.1
81.4
33.3
59.3
92.6
39.1
34.2
73.3
49.8
44.7
94.5
51.1
36.4
87.5
38.7
63.7
102.4
84.4
38.4
122.8
50.4
30.3
80.7
32.8
27.5
60.3
65.1
39.1
104.2
36.7
39.3
76.0
236.3  276.1
 79.0   51.6
315.2  210.4

251.1  210.4
 77.0   74.5
328.1  284.9
 236.1   207.9
  33.8    66.8
1270.0   274.7

 167.7   150.6
  48.9    59.0
 216.6   209.5
(Concentrations in Milligrams/Cubic Meter)
     Semivolatile     56.0   27.6      56.4    36.6
     Gravimetric    262.0  136.1     316.4   161.5
     Total EOM     319.0  163.8     372.9   198.1
                Table 2.  Percent of Sample Collected by
                    XAD-2 and Paniculate Filter
Site
Ambient (EGP, Fire, Out)
                        XAD-2
                          76
Indoor (With and Without Woodstoves)    94

Stack                                 41
                       Particula*6

                          24

                           6

                          59
                              826
-------
0)
                                    GRAVIMETRIC EOM
                                    SEMI-VOLATILE EOM
                 i I i I i i i i ( i 11 r i i
         Weekend      Weekday
                 Day
                           I I I t I I II I I I I I I I I I It I 1TI
                                Weekend      Weekday
                                       Night
   e 1:  Typical  ambient (fire)  concentrations
                                       GRAVIMETRIC EOM
                                       SEMI-VOLATILE EOM
        Weekend
               Day
                 Weekday
Weekend... ..Weekday
       Night
     2:  Typical indoor  (without  woodstove)  concentrations
                           &
                                    •  GRAVIMETRIC EOM
                                    *  SEMI-VOLATILE EOM
                         t-
   Weekend Q     Weekday

3: Stack concentrations
                                   Weekend...  .. Weekday
                                          Nignc
                            827
-------
GC/MS ANALYSIS OF WOODSTOVE EMISSIONS AND AMBIENT SAMPLES FROM A WOOD
SMOKE IMPACTED AREA

R. S. Steiber and J. A. Dorsey
U. S. Environmental Protection Agency
Air and Energy Engineering Research Laboratory
Mail Drop 62B
Research Triangle Park, North Carolina  27711

ABSTRACT

        Gas Chromatography/Mass Spectrometric (GC/MS) analyses have been conducted on an i
set of samples acquired to assess the impact of wood stoves on ambient air quality. The sample set was
acquired in Boise, Idaho, from November 1986 through February 1987. The set includes stove emissions
and ambient air samples. Sampling was conducted during the periods of 7:00 a.m. to 7:00 p.m. (DAY) an"
7:00 p.m. to 7:00 a.m. (NIGHT) using samplers with both particulate filtration and XAD-2 sorbent for
semi-volatile collection.
        Stove emission and ambient air samples are dominated by the presence of methoxybenzenes
are the products of thermal decomposition of the lignin in the wood. The polycyclic aromatic hydrocarbon8
(PAHs)  present in these samples are at a much lower concentration than the methoxybenzenes.

BACKGROUND

        The purpose of this paper is to present data from the GC/MS analysis of samples from the
Integrated Air Cancer Project's Boise winter study. The samples were of two basic types: the
dichloromethane extracts of the particulate catch on a Teflon filter and a followup canister containing
approximately 175 g of XAD-2 resin. The manner in which these samples were acquired and extracted is
described elsewhere (1,2), and this paper will present analytical results only. In excess of 600 samples of
this type were taken during the course of the Boise study. This paper will concentrate on the results fro111
one weekend only, hoping by that method to present a snapshot of the kinds of compounds emitted by a
typical western wood stove during a single 48 hour period and their distribution throughout an urban and
suburban area.

        The samples described in this paper were acquired at four separate sites in Boise, Idaho, beginning
at 7:00 a.m. on Saturday, November 22, 1986, and concluding at 7:00 a.m. on Monday, November 24.
After extraction, the samples were composited according to the period in which they were taken: DAY
(7:00 a.m. to 7:00 p.m., Saturday and Sunday) or NIGHT (7:00 p.m. to 7:00 a.m., Saturday and Sunday)-
This resulted in a total of 16 separate daytime and nighttime samples for the weekend (2 composite filter
samples and 2 composite XAD-2 samples for each site).

        The sites were as follows: the chimney of a residential house with a wood stove (STACK); a
nearby outdoor location (OUT); Elm Grove Park, a playground located in the same general residential area
(EGP); and the roof  of a fire station (FIRE) located near a heavily travelled road. Indoor air samples were
also taken both from the house with  a wood stove and a nearby house without a wood stove, but the resul'
from these samples are not discussed in this paper.  Concentrations for each site are given in Table 1.
                                             828
-------
                                           TABLE 1
                     CONCENTRATIONS IN MICROGRAMS/CUBIC METER
                               DICHLOROMETHANE EXTRACTS
                                WEEKEND 2, 11/22/86-11/23/86

                                                              DAY           NIGHT
                EGP            XAD-2                        48.4            55.7
                                PARTICIPATE                16.7            24.2

                OUT            XAD-2                        55.1            72.8
                                PARTICULATE                16.5            25.0

                FIRE           XAD-2                        57,8            72.8
                                PARTICULATE                12.6            21.0

                STACK         XAD-2                      93,468         48,256
                                PARTICULATE             155,910         86,364
 p.  .   After extraction with dichloromethane and concentration to 5 ml, the samples were analyzed on a
   mgan Model 5100 Quadrapole GC/MS. A 30 m DB-5 microbore capillary column was used, and the
   Perature program was 40° C for 3 minutes followed by a 10°C per minute ramp to 280°C. Before
 u losing the results of this work, it is important to note certain limitations of this instrument.  In the
 Of ih  ?^' that P°rti°n of tJie column which extends into the source is unheated. This makes resolution
 °ff th ^8ner molecular weight polycyclic aromatic hydrocarbons (PAHs) difficult since they tend to bleed
   toe end of the column rather than elute as coherent peaks. To overcome this problem, the interface oven
 6^ heated to 285°C and the manifold heater raised to near its maximum safe working temperature (120°C).
 e f*°> no consistent results could be obtained for PAHs with ring structures of six members or more;
  *• °enzo(ghi)perylene (MW276).

 STACK SAMPLES

 are    Stack samples were acquired using a dilution sampling system in which both the filter and XAD-2
   at ambient temperature. The stack samples provided the greatest concentrations of extractable organic
 2aterial (EQM) of 0.1 - 0.2 g/m3. The filter catch was particularly rich, nearly doubling that of the XAD-


 n^    The majority of the identified compounds fell into six basic categories:  aliphatic hydrocarbons,
  iji   ^kanes and alkenes; monoaromatic hydrocarbons, particularly the alkyl benzenes; condensed-ring
     , s including many of the classic PAHs; monoaromatic oxygenated species as aldehydes, ketones, and
      ; oxygenated monoaromatics such as phenol, methoxy phenol, and methoxy benzenes; and
          PAHs. Also present were a number of nitrated species.
       The cut between species caught on the filter and those found in the XAD-2 was largely a function
       g point with the more volatile species being captured in the XAD-2. There was, also, a certain
  y tlt of species differentiation between the two cuts. The alkanes, alkenes, alkyl benzenes, and single-
  fh- n monoaromatics tended to end up in the XAD-2 portion. The filter catch, on the other hand, was
    jjated by the multi-oxygenated species and the PAHs, particularly those with three-, four-, and five-
       ring structures.

       Several differences were apparent between the daytime and nighttime stack samples (See Figure 1).
    L°ncentrations 'n lhe daytime sample are more than twice those of the nighttime sample. This was
       ^ue to tne manner 'n which the wood stove was used in this particular household. At night, for
    fC> tlle fire was banked around bedtime and slowly allowed to die during the hours of early morning.
       mation of f AHs is closely related to conditions within the firebox, the cooling of the wood fire
      nd to inhibit the chemical reactions necessary for their production. The reverse process seems to
  e
due /^n at work in the case of the alkanes and alkenes. These increased during the night, again probably
f|frth  lhe dy.ing of lne fire' As tne fire coolcd. more of lnese compounds would simply boil off without
    r chemical change. Otherwise the daytime and nighttime samples are quite similar, except perhaps for



                                            829
-------
 the presence of more nitrated compounds (X NITRO) during the daytime.  Their formation would again be a
 function of the temperature of the fire.

         In all of these stack samples, both day and night, the dominant species were the oxygenated
 monoaromatics. The presence of these compounds is consistent with the thermal destruction of lignin.
 Lignin constitutes up to 30% of all woods (3) and may be described as an  almost infinite network of
 branch-chain polymer molecules. Other constituents, such as the celluloses and hemicelluloses, combine
 with lignin to give wood its rigidity in a structure that resembles reinforced concrete (the lignin forming &e
 core, the celluloses the supporting mesh).

         While the precise structure of the lignin polymer has not been fully determined, it can be said to
 consist chiefly of two structural units. These are a guaiacyl nucleus in a guaiacylpropane skeleton and a
 syringyl nucleus in a syringylpropane skeleton (4). (Sec Figure 2.)

         The fracture of these structures from their networks during combustion would lend to produce a
 wide variety of methoxy phenols, methoxy benzenes, and alkyl benzenes; and this is precisely what app6318
 to have happened in the stack samples under discussion. In addition, the broad range of temperatures
 available in a wood fire would provide the conditions for the transformation of some of these structures in'0
 naphthalenes, pyrenes, and other condensed ring aromatics.  Table 2 presents a typical list of compounds
 found in the stack sample.

 AMBIENT SITES

        The ambient sites were chosen for the purpose of measuring the impact of wood smoke on the
 outside air.  Two of them (EGP and OUT) were located in a suburban area, while the FIRE site was situate"
 in a place where auto exhausts would also have an impact. Insofar as the total cxtractable organic matter
 (EOM) catch was concerned, the compounds identified were consistent with the ambient circulation of ^^
 stove effluents. This was as much true at the FIRE site as it was at the EGP and OUT locations.

        In terms of sheer numbers of compounds, the daytime OUT XAD-2 sample was dominated by
 aliphatic hydrocarbons, alkylatcd monoaromatics, and a homologous series of substituted naphthalenes,
 with few other species appearing. During the evening hours, this group broadened somewhat to include
 such oxygenated monoaromatics as methyl bcnzofuran and dimethyl bcnzaldchydc.  The nighttime filler
 catch was dominated equally by oxygenated monoaromatics (particularly phenols) and PAH. This last
 group included chrysene, retene, bcnzo(e)- and bcnzo(a)pyrcne, and bcnzo(k)fluoranthene.

        The classes of compounds found at the EGP site bore a close resemblance to those identified  in l"
 OUT samples, except that the higher molecular weight PAHs were absent  from the nighttime  filter catch-
 Once again, the alkylated monoaromatics dominated the XAD-2 extracts, while oxygenated aromatics *ere
 heavily represented in the filter catch.

        The distribution of compounds at the FIRE site was similar to that found at both the  EGP and ^
 OUT sites.  In fact, when plotted side by side, as in Figures 3 and 4, they appear to track each  other
 remarkably  well.  Although it was expected that auto exhaust would have a significant impact at this sit6-
 there appears to have been no greater contribution than at the other sites.
        We have pointed out the similarities among the EGP, OUT, and FIRE samples. The
of EOM in these same samples also correlates well with what was being emitted from the stack. The
classes of compounds are found in roughly the same concentrations relative to the size of the sample c
Of course, the ambient sites were impacted by the effluent from many wood burning sources and not just
the one  selected for examination during this particular weekend. What this shows, however, is that
whatever their source, there is an easily recognizable fingerprint for wood smoke effluents, and that wood
burning sources (at least in  the area surveyed) seem to produce the same range of compounds.

FINGERPRINT COMPOUNDS FOR WOOD SMOKE

        One of the goals of the analytical scheme was to identify fingerprint compounds for wood
Some investigators have proposed l-methyl-7-(l-methylethyl)-phenanthrene or retene as such as fi
While retene was found in the stack and a few of the ambient samples, it was not a suitable species
purpose. Any fuel that is rich in aromatics will produce condensed-ring aromatics under the right
conditions, so there is nothing about retene that unconditionally ties it to the burning of wood.
                                              830
-------
        Another class of compounds that looked promising were the terpenes.  These arc a group of
        CIQ cyclo- and bicyclo-alkancs and alkcnes. They are found naturally in woody plants, particularly
 v j!  .s' Among ^str number they include pinene, the characteristic odor of pine trees; carene, the
 CM      ec' C0'or'n8 agent in carrots, peaches, and many woods; and thujene, the characteristic odor of
  vtor. Since western white pine is the fuel of choice in most Boise wood stoves and, since significant
 founts of terpenes tend to boil off a wood fire without undergoing further chemical change, it was decided
 p1 this might be a class of compounds to look for.  Multiple terpenes turned up in all of the samples
  scusscd in this paper, and there can be little doubt that one source was the wood stove effluent. However,
 ^ a Universal indicator for wood smoke they have certain drawbacks. One such drawback is their ubiquity.
        are w^e'v use<* 3s deodorants and flavoring agents and, in the electronics industry, as solvents. In
 ..     , the source of pinene in an ambient sample might not be wood smoke at all but a nearby stand of
 lir trees.

 .      The most reliable indicator for the presence of wood smoke turned out to be the mcthoxy phenols.
   Mentioned above, the thermal destruction of woody lignins tends to produce a wide variety of mcthoxy
 in n?-°ls and mcmoxy benzenes.  Compounds of these kinds appeared in all the ambient samples discussed
      PapCr and> when lhe results of both lne XAD'2 and lhe filter extract samples are combined, their
        correlate well with what was found in the stack. Figure 5 presents these results in bar graph form.
   e single oxygen aromatics would include the phenols and methoxy benzenes; the two-oxygen aromatics,
 ,  methoxy phenols and dimethoxy benzenes; the three-oxygen, the dimcthoxy phenols, the trimethoxy
 ^ nzenes; and so forth. The pattern visible in these four samples is striking evidence of the transport of
  9od smoke among the three sites. While the use of statistics based on small numbers of samples can be
 and K ing> il is interesting &at tne correlation coefficient for the STACK and the EGP samples was 0.89
 vj it  same numocr f°r the STACK and the FIRE samples was 0.87, These results are consistent with
   **• done by Hawthorne et al. (5) in connection with wood smoke derived methoxylated phenols.

 INCLUSION

        o Differences were noted between daytime and nighttime stack samples, including a higher
        concentration of PAH and a lower concentration of aliphatic hydrocarbons in the daytime
        sample. Monoaromatics and methoxylated benzenes and phenols appeared to remain fairly
        constant.

        o Both stack and ambient samples were rich in monoaromatics and methoxylated benzenes and
        phenols.

        o Wood smoke appeared to be a major contributor to the EOM
        catch at the FIRE and EGP sites.

        o The methoxylated benzenes and phenols are the best class of compounds to use as wood smoke
        tracers.

                                       REFERENCES
 1,
       R. Martz, D. Natschke, "Large Scale Cleaning of XAD-2 Sorbent Resin for Air Sampling,"
       Proceedings of the 1988 EPA/APCA Symposium on Measurement of Toxic Air Pollutants.
I.
      R. McCrillis, P. Burnet, "Effects of Operating Variables on Emissions from Woodstoves,"
      Proceedings of the 1988 EPA/APCA Symposium on Measurement of Toxic Air Pollutants.
3.
      C. Libby, Pulp and. Paper Science ajod Technology. Volume 1 , Chap. 5, pp. 82-107 (McGraw-
      Hill, New York, 1962).
4
      Ibid.
S.
      S. Hawthorne, M. Krieger, D. Miller, "Methoxylated Phenols as Candidate Tracers for
      Atmospheric  Wood Smoke Particulates," Proceedings of the 1988 EPA/APCA Symposium on
      Measurement of Toxic Air Pollutants.
                                            831
-------
        TABLE 2: MAJOR COMPOUNDS DETECTED IN STACK SAMPLE WE2 DAY

PERCENT          COMPOUNDS

                ********** ALKANES, ALKENES, CYCLOS **********

0.76   DIMETHYL HEXADIENE, C8 H14, MW 110
0.56   METHYL HEPTANE, C8 HIS, MW 114
0.14   C8 CYCLO-ALKENE, C8 H14, MW 110
0.37   SUBSTITUTED C8 BICYCLO-ALKANE, C8 H14, MW 110
1.41   PINENE.C10H16, MW 136
0.08   SUBSTITUTED CIS ALKENE OR CYCLO-ALKANE, CIS H30, MW 210
0.10   PENTADECANE, CIS H32, MW 212
0.06   SUBSTITUTED C16 ALKANE, C16 H34, MW 226
0.07   HEPTADECANE, C17 H36, MW 240
0.09   SUBSTITUTED C17 ALKANE, C17H36, MW 240
0.01   C17-18 ALKENE OR CYCLO-ALKANE

          ********** OXYGENATED ALKANES, ALKENES, CYCLOS **********

0.01   ACID ESTER, Cx Hx 02
2.48   2-FURANCARBOXALDEHYDE, C5 H4 O2, MW 96
1.08   FURAN-3-ALDEHYDE, C5 H4 O2, MW 96
1.73   l-ACTETYLOXY-2-PROPANONE, C5 H8 O3, MW 116
0.32   FURANYL ETHANONE, C6 H6 O2, MW 110
0.21   HYDROXY-DIMETHYL-BUTANONE, C6 H12 O2, MW 116
0.01   2,5-CYCLOHEXADIENE-l, 4-DIONE, C6 H4 O2, MW 108
0.01   2-METHYL-2-PROPENOIC ACID, ETHYL ESTER, C6 H10 O2, MW 114
0.00   2-METHYL-2, 5-CYCLOHEXADIENE-l, 4-DIONE, C7 H6 O2, MW 122
0.59   5-METHYL-3-HEXEN-2-ONE, C7 H12 O, MW 112
0.01   UNIDENTIFIED CHO, C7 H10 O, MW 110
0.02   UNIDENTIFIED CHO, C8 H8 O3, MW 152
2.06   1-(1-CYCLOHEXEN-1-YL)-ETHANONE, C8 H12 O, MW 124
0.05   SUBSTITUTED ALCOHOL, CIO H14 O2, MW 166
0.01   CHO, Cl 1 H14 O4, MW 210
0.07   UNIDENTIFIED CHO, C13 H28 O. MW 200

                    ********** MONOAROMATICS **********

5.16   TOLUENE, C7 H8, MW 92
0.10   1,4-DIMETHYL BENZENE, C8 H10, MW 106
2.57   1,2-DIMETHYL BENZENE, C8 H10, MW 106
2.99   1,3-DIMETHYL BENZENE, C8 H10, MW 106
1.60   ETHENYL BENZENE, C8 H8, MW 104
2.95   l-ETHENYL-3-METHYL-BENZENE, C9 H10, MW 118
0.07   SUBSTITUTED C9 BENZENE, C9 H12, MW 120
0.50   l-ETHYL-3-METHYL-BENZENE, C9 H12, MW 120
0.15   PROPYL BENZENE, C9 H12, MW 120
3.55   l-ETHYNYL-4-METHYL-BENZENE, C9 H8, MW 116
0.36   (METHYLENECYCLOPROPYL)-BENZENE, CIO H10, MW 130
2.75   1-ETHYL-l, 4-DIMETHYL BENZENE, CIO H14, MW 134
0.63   METHYL (1-METHYLETHENYL)-BENZENE, CIO H12, MW 132
0.31   (l-METHYL-2-CYCLOPROPEN-l-YL)-BENZENE, CIO H10, MW 130
0.18   SUBSTITUTED BENZENE, Cll H14, MW 146
0.89   U-BIPHENYL, C12 H10, MW 154
0.11   2-METHYL-l, 1-BIPHENYL, C13 H12, MW 168
0.06   3,3-DIMETHYL-l, 1-BIPHENYL, C14 H14, MW 182
0.01   SUBSTITUTED CIS BENZENE, C15 H16, MW 196
0.05   U-METHYLENEBIS (3-METHYL-BENZENE), CIS H16, MW 196
0.01   (ETHYLPHENYL)-ETHANE, C18 H22, MW 238
0.05   (I-BUTYLOCTYL) BENZENE, CIS H30, MW 246
                                    832
-------
               ********** OXYGENATED MONOAROMATICS **********
   - -   PHENOL, C6 H6 O, MW 94
  JjH   AROMATIC CHO, C6 H6 O2, MW 110
  u'°3   CHO.C6H602.MW110
       METHYL PHENOL, C7 H8 O, MW  108
       3-METHYL PHENOL, C7 H8 O, MW 108
       4-METHYL PHENOL, C7 H8 O, MW 108
       2-HYDROXY-BENZALDEHYDE, C7 H6 O2, MW 122
  «'   2-METHYL-PHENOL, C7 H8 O, MW 108
  W   DIMETHOXY PHENOL (ISOMER), C8 H10 O3, MW 154
  ,'"   3,4-DIMETHOXY PHENOL, C8 H10 O3, MW 154
  0'°J   2.5-DIMETHYL PHENOL, C8 H10 O, MW 122
  0-'«   HYDROXY METHOXY BENZOIC ACID (ISOMER), C8 H8 O4, MW 168
  .u   ETHYL OR DIMETHYL BENZENEDIOL, C8 H10 O2, MW 138
      4-HYDROXY-3-METHOXY BENZALDEHYDE, C8 H8 O3, MW 152
      1-(2,4-DIHYDROXYPHENYL)-ETHANONE, C8 H8 O3, MW 152
       3,4-DIMETHYL PHENOL, C8 H10 O, MW 122
      2,6-DIMETHOXY PHENOL, C8 H10 O3, MW 154
      0-ETHYNYL-PHENOL, C8 H6 O, MW 118
      3-HYDROXY-4-METHOXY BENZALDEHYDE, C8 H8 O3, MW 152
      ETHYL PHENOL, C8 H10 O, MW 122
      3-ETHYL PHENOL, C8 H10 O, MW 122
      4-HYDROXY-3-METHOXY BENZOIC ACID, C8 H8 O4, MW 168
      ETHYL BENZENEDIOL, C8 H10 O2, MW 138
      4-ETHYL-l, 3-BENZENEDIOL, C8 H10 O2, MW 138
 Q' '   HYDROXY METHOXY BENZENEACETIC ACID, C9 H10 O4, MW 182
 5' "   l-(2,6-DIHYDROXY-4-METHOXYPHENYL)-ETHANONE, C9 H10 04, MW 182
 Q'^   4-ETHYL-2-METHOXY PHENOL, C9 H12 O2, MW 152
 0 0?   SUBSTITUTED C9 PHENOL, MW 132
 0.'   K2-HYDROXY-5-METHYLPHENYL)-ETHANONE, C9 H10 O2, MW 150
 0'^   PHENYL PROPENOIC ACID,  C9 H8  O2, MW 148
 O'Q,   SUBSTITUTED METHOXY BENZENE, C9 H12 O. MW 152
 O'fi;   (2-PROPYNYLOXY)-BENZENE, C9 H8 O, MW 132
 3*0r   SUBSTITUTED PHENOL, CIO H14 O, MW 150
 fid   2-METHOXY-4-PROPYL-PHENOL, CIO H14 O2, MW 166
 0','   3,5-DIETHYL PHENOL, CIO H14 O, MW 150
  '°  H3,4-DIMETHYLPHENYL)-ETHANONE, CIO H12 O, MW  148*
     1-(2-HYDROXY-5-METHOXY-4-METHYLPHENYL)-ETHANONE, CIO H12 O3.MW180
      PROPYL GUAIACOL, CIO H14 O2, MW 166
 0/vJ  H4-HYDROXY-3-METHOXYPHENYL)-2-PROPANONE, CIO H12 O3, MW 180
 O'jJ  "f-O.l-DIMETHYLETHYLJ-BENZENEDIOL, CIO H14 O2, MW 166
 00T  2-METHOXY-4-(2-PROPENYL)-PHENOL, CIO H12 O2, MW 162
 o0*  METHOXY PROPYL PHENOL, cio HW 02, MW 166
 Ooi   1-(2,4,6-TRIHYDROXY-3-METHYLPHENYL)-l-BUTANONE, Cl 1 H14 O4, MW218
 Q0   DIMETHOXY (PROPENYL) PHENOL, Cl 1 H14 O3, MW 194
 o.0J   PHENYL HEXADIYNONE, ci2 HS o, MW 168
 0.08   DIMETHOXY TETRAMETHYL BENZENE, ci2 HIS 02, MW 194
 0,0l   ('.I-BIPHENYL)-4-CARBOXALDEHYDE, C13 H10 O, MW 182
 Oi04   DIPHENYL OXIRANE, C14 H12 O, MW 196
 O.QQ   4-(2-PHENYLETHENYL)-PHENOL, C14 HI2 O, MW 196
 o.o3   SUBSTITUTED cis PHENOL, cis H24 o, MW 220
 0.0?   PHENYL (TERT-BUTYL-HYDROXY-PHENYL)-ETHANE, CIS H22 O, MW 254
     4-HYDROXY-3,5-DIMETHOXY BENZALDEHYDE, C19 H10 O4, MW 182

         ********** POLYCYCLIC AROMATIC HYDROCARBONS **********
 8,13
 l.9o   ^HTHALENE, cio HS.MW  128
2.76   2-METHYL NAPHTHALENE, Cll H10. MW 154
o.l6   -METHYL NAPHTHALENE, ci i mo, MW 142
0.31   '-ETHYL NAPHTHALENE, C12 H12, MW 156
     2,3-DIHYDRO-4,5,7-TRIMETHYL INDENE, C12 H16, MW 160
                                 833
-------
0.66   2-ETHENYL NAPHTHALENE, C12 H10, MW 154
0.36   2-ETHYL-NAPHTHALENE, C12 H12, MW 156
0.45   1,8-DIMETHYL NAPHTHALENE, C12 H12, MW 156
0.15   1,2-DIHYDROACENAPHTHYLENE, C12 H10, MW 154
0.21   2,6-DIMETHYL NAPHTHALENE, C12 H12, MW 156
3.85   ACENAPHTHYLENE, C12 H8, MW 152
0.48   9H-FLUORENE, C13 H10, MW 166
0.06   TRIMETHYL NAPHTHALENE, C13, H14, MW 170
0.11   1 H-PHENALENE, C13 H10, MW 166
0.03   METHYL ACENAPHTHALENE, C13  H10, MW 166
1.84   PHENANTHRENE, C14 H10, MW 178
0.22   1-(I,1-DIMETHYLETHYL)-NAPHTHALENE, C14 H16, MW 184
0.10   PHENANTHRENE, C14 H10, MW 178
0.10   2-METHYL-9H-FLUORENE, C14 HI2, MW 180
0.24   ANTHRACENE, C14 H10, MW 178
0.10   9-METHYL-9H-FLUORENE, C14 H12, MW 180
0.05   METHYL FLUORENE (ISOMER), C14 H12, MW 180
0.03   4H-CYCLOPENTA (OEF) PHENANTHRENE, C15 H10, MW 190
0.03   I-METHYL ANTHRACENE, CIS H12, MW 192
0.01   4-METHYL PHENANTHRENE, C15 H12, MW 192
0.01   2,3-DIMETHYL-9H-FLUORENE, C15 H14, MW 194
0.02   METHYL PHENANTHRENE
0.01   METHYL ANTHRACENE, C15 HI2, MW 192
0.03   2,5-DIMETHYL PHENANTHRENE, C16 H14, MW 206
0.03   FLUORANTHENE, C16 H10, MW 202
0.01   DIMETHYL PHENANTHRENE, C16 H14, MW 206
0.01   2-PHENYL-NAPHTHALENE, C16 HI7, MW 204
0.03   PYRENE, C16 H10, MW 202
0.01   4,5-DIMETHYL PHENANTHRENE, C16 H14. MW 206
0.00   1-(PHENYLMETHYL)-NAPHTHALENE, C17 H14, MW 218
0.03   RETENE (l-METHYL-7-ISOPROPYL PHENANTHRENE), C18 H18, MW 234

           ********** OXYGENATED POLYCYCLIC AROMATICS **********

0.05   SUBSTITUTED NAPHTHOATE, Cx Hx O2
1.30   BENZOFURAN, C8 H6 O, MW 118
0.06   1,2,3,4,4A,9,10,10A-OCTAHYDRO-1,4A-DIME:1-PHENANTHRENE
      CARBOXALDEHYDE, CIO H28 0, MW 284
0.09   2-METHOXY-3-BENZOFURANCARBOX ALDEHYDE, CIO H8 O3, MW 176
0.06   4-METHOXY-NAPHTHALENOL, Cl 1 H10 O2, MW 174
3.19   DIBENZOFURAN, C12 H8 O, MW 168
0.13   4-METHYL DIBENZOFURAN, C13 H10 O, MW 182
0.12   9H-FLUORENONE, C13 H8 O, MW 180
0.14   METHYL DIBENZOFURAN, C13 H10 O, MW 182
0.01   PHENANTHRENOL, C14 H10 O, MW 194
0.02   1,2,3,4,4A,9,10,10A-OCTAHYDRO-1,4A-DIMETHYL PHENANTHRENE
      CARBOXYLIC ACID, C21 H30 O2, MW 314

                   ********** HETERO COMPOUNDS **********

0.03   CHLOROAROMATIC, MW 256
0.02   1,1-SULFONYLBIS (4-CHLORO)-BENZENE
0.18   UNIDENTIFIED CHN
0.08  N,N-DIMETHYL-4-(((l-METHYLETHYL)lMINO)METHYL)-BENZAMINE,C12 H18 N2,
      MW190
0.01   CHN.C14H13N, MW 195
0.04   1,4-BIS (4-METHYLPHENYL) SULFONYL-PIPERAZINE, Cl8 H22 O4 N2 S2, MW394

               Note: Positions of substituted radicals based on best computer fit
                                   834
-------
 EpFECTS OF OPERATING VARIABLES ON EMISSIONS FROM WOODSTOVES
       C. McCrillis
 j*r and Energy Engineering Research Laboratory
  •S. Environmental Protection Agency
 Research Triangle Park, North Carolina 27711
     G. Burnet
     Environmental Services, Inc.
 J0950 SW 5th Street, Suite 160
 Beaverton,  OR  97005

 Abs tract

     During the winter of 1986-87,  the U.S. Environmental  Protection
 ?ency (EPA) conducted an emission  measurement  program in  Boise,  ID, as
 •art  of the Integrated Air Cancer Project  (IACP).   This program was
 esigned to identify the  potential  mutagenic impact  of residential wood
 lrnlng on  ambient and indoor  air.   One facet of  this  field
 ainPling effort involved  obtaining  emission samples  from chimneys
 *6rving wood burning appliances  in  Boise.   As a companion  to  the  field
 °ufce sampling,  a parallel  project was undertaken  in  an instrumented
 °°dstove test  laboratory to quantify  woodstove emissions  during
  erations  typical of  Boise  usage.

 ^    Two woodstoves were  operated in a test  laboratory over a  range  of
 Urnrates,  burning either eastern oak  or white  pine  from the Boise,  ID,
 tea*   A conventional  stove, manufactured  in the Boise area, was  tested
 t  altitudes  of  90 and 825 m.  A catalytic stove was tested only at
  e high altitude  facility.  All emission  tests were started with
        a fire in a  cold  stove using black and white newsprint
,*lssions were collected using the wood  stove dilution sampling system
^ WSDSS) for a continuous period of about 8 hours, encompassing start-up
  1 several wood additions.  The  results showed wide variability
        due primarily to the difficulty  in duplicating conditions
      start-up.  Total WSDSS emissions showed the expected inverse
f Delation with burnrate for the conventional stove and nearly flat
c r the catalytic stove.  While there appeared to be little or no
Delation of total WSDSS emissions with altitude, the sum of the 16
c ^nuclear aromatic hydrocarbons (PAHs) quantified showed a direct
 Delation with altitude: higher PAH emissions at the higher altitude.
    JXiring the winter of 1986-87, the U.S. Environmental Protection
    y (EPA) conducted an emission measurement program in Boise, ID, as
    of the Integrated Air Cancer Project (IACP).  This program was
    ned to identify the potential mutagenic impact of residential wood
    n6 on ambient and indoor air.  One facet of this field sampling
   rt involved obtaining emission samples from chimneys serving wood
       appliances in Boise.  As a companion to the field source
       ,  a parallel project was undertaken in an instrumented
      ve  test laboratory to quantify woodstove emissions during
j ^rations typical of Boise usage.  The results from these
  °ratory source tests are the subject of this paper.
                                  835
-------
  Experimental Design

       Nearly all of  the woodstove  data in the  literature  have  been
  obtained in a laboratory setting  with the start  of  an  emission  test
  occurring after the fire was  lit  and  the appliance  brought up to
  operating temperature.  In  many cases,  these  tests  also  used
  dimensional lumber  as  fuel  as  specified in various  regulatory
            f 1  O  O \              r                     O     J
  procedures^1 >^ »J' .   In a moderate winter climate such  as found
  in Boise, ID, it is common  practice for woodstove users  to kindle a
  new fire in a cold  stove in the morning.   This fire is often allowed
  to die  out  during the  day when heating  demand falls.   A  new fire is
  kindled during  the  early evening  which  is  then stoked  for the night
  and,  oftentimes, burns  out  before the residents rise the next
  morning.   Since  the  objective  of  this work was to obtain emission
  samples  under operating  conditions similar to those observed in
  Boise,  ID,  it was decided that each emission test would start with
  lighting  a  fire  in  a cold stove.

        To limit  the number  of  tests required to obtain statistically
  valid results, the number of operating variables to be Investigated was
  limited  to  four:  fuel type (wood species), burnrate, stove type, and
  altitude.  Each of these variables was investigated at two levels.

      Fuel ---------- eastern oak and white pine from Boise, ID,  area
      Burnrate ------ high and low values
      Stove type ---- conventional airtight stove manufactured in Boise,
                    ID,  area and a  catalytic stove
      Altitude ------ 90 and 825  m

 Eastern oak was burned  at the  90 m elevation to  provide a tie  to
 earlier IACP source  laboratory tests^'.  The  tests  at  825  m were to
 provide data at an altitude  equivalent to the  residential area studied
 in Boise, ID.  Tests on the  catalytic  stove were  Included to gain some
 understanding of the changes to be expected in source  emissions  as
 emission technology  stoves become  more common  because  of  recent  EPA
 regulations'*' •

      To  prepare  for  an  emission test,  the stove and  flue  pipe were
 brushed  and  vacuumed clean.  A  pretest fire was then lit  and burned
 several  hours  at  the conditions of wood  species and burnrate planned  £°
 the next test.  The  pretest  fire was allowed to burn out, and the stove
 cooled to room temperature.  To start  a  test, several newspaper  balls
 and kindling wood were  placed in the stove.  All sampling equipment
 started  when the  paper  was ignited.  The  stove loading  door was  left
 open for 5-10 minutes until a good fire was established.  At this
 additional wood was  loaded into the stove and the door  closed.   Each
 emission  test  lasted for  about  8 hours; wood was added  periodically as
 needed to  maintain the desired  overall burnrate for the test.
     All emission samples were collected with the wood stove
sampling system (WSDSS).  This system, described In detail elsewhere
removes a sample directly from the flue exit and dilutes the sample wi
cleaned ambient air simulating plume formation.  The cooled and
diluted sample then passes through a Teflon coated filter and XAD-2
adsorbent resin.  During the tests, the filter was changed anytime the
pressure drop across it became excessive.  In all of the tests it was
necessary to change the filter several times over the course of an 8
hour burn.  Excessive pressure drop was usually encountered within a
short time after the addition of  fresh fuel to the stove.

                                   836
-------
      The  WSDSS  samples  recovered  at  the  end  of  each  test  consisted  of
   116  filter(s),  XAD-2, and  probe wash.  The probe wash  consisted  of
  J-Parate  dichlororaethane and raethanol  rinses.   Representative samples  of
   fje  wood  burned and the ash were  also  collected for  elemental analysis.
  Jje  WSDSS filters were weighed and then  extracted with dlchloromethane.
   [>s  XAD-2 was similarly extracted.   These separate extracts and  the
   Jchloromethane probe wash were analyzed separately  for total organic
  T*88 in two steps.  The semivolatile mass was quantified by gas  chroma-
   °Sraphy, and the condensible mass, gravimetrically .  The methanol probe
   ash was analyzed gravimetrically.  Selected PAHs were quantified by
      pressure liquid chromatography(6) .

      A. 1 1/min slipstream of the diluted sample was removed from the
       upstream of the filter for aldehydes analysis.   A 5 1/min slip-
        was removed between the  filter and XAD-2 cartridge for
  ydrocarbon analysis.
      The  following  discussion summarizes  the results  of  analyses
  oupleted to  date on the  WSDSS samples.   Still  to  be  completed are the
  :°assays.  These and other  data collected,  such as the  hydrocarbon and
 niemental analyses,  will  be  reported  later.   The aldehyde  samples  did
   *  comply with quality assurance requirements  and will  not  be reported.

 g    Figure 1 presents WSDSS  emission results for  all valid  test burns.
   ch  bar  is composed  of three  parts:  the semivolatile,  condensible, and
 f ^extract able fractions.  For most of the test burns the  nonextractable
 I action  is larger than usually  seen in woodstove  samples.   This may be
  *rgely due to the cold stove  start employed in these tests, whereas
 /evious data were taken during  hot start tests only.  During  start-up,
   a   is much higher than at other times which may have  carried more ash
     cles (including bits of newspaper) up the flue.  The ratio of
    volatile to condensible fraction ranged from 0.13 to 2.2 with an
  erage value of 0.35 which is in general agreement with earlier
 /g||ults<5,7).  Note the sraall variability between some replicate burns
 (Sn  ~* and SOLL~2> compared to the large variability between others
     ~ 1 and SOLH-2)'  AS n°ted previously, this variability was
            (but certainly not welcomed!) as  a result of the cold start
        of  the  test  program.

 h    Figure  2 presents  the  same  WSDSS emission results  as  a function of
 |j rnme.  There  are three  data  points for  each burn  plotted at  the  same
 v !;nrate.  The  circumscribed numbers  are  the  total  train emission
        The diamonds represent  the condensible emissions,  while the
        are the semi volatile emissions.   The  total  minus the
 e *lvolatile and  condensible fractions equals  the nonextractable
 IJL 8s*on rate.  With exception of Burn 2, the  conventional  stove data
  **  the expected trend:  high emissions  at low burnrates decreasing
       and leveling out at  high burnrates.  It is  of interest to note
tfc * ^e condensible fraction emission rate trend is similar; however,
&ue 8emi volatile emission rate is relatively constant with burnrate.
6j n 2 is an anomaly because the test was terminated early due to an
-------
available, It appears that the catalytic stove emission characteristic
is similar to that for other models of this technology i.e., an
increasing emission rate with burnrate.

     Figures 3 and 4 illustrate the relationship between the PAH
emission rates and the burnrate.  The sum of the 16 PAHs and naphthalene
showed some correlation (r^ = 0.48 and 0.58, respectively).  The
correlation coefficients for pyrene and benzo-a-pyrene were much weaker
(0.23 and 0.04, respectively).  It is important to note that the PAH
burnrate relationship for all the data seems to be direct as compared to
the inverse relationship between total WSDSS emissions and burnrate seat1
for the conventional stove which constitutes the bulk of the data.  This
indicates that for these tests as total emission rate decreased with
increasing burnrate, the percent of the emissions constituting the PAH
fraction increased.

     An analysis of variance performed on these data showed few statis*
tically significant correlations (main effects) because of the wide
variability.  One of the main effects identified thus far is the
influence of altitude on PAH emissions.  The statistical analysis
shows that increasing altitude from 90 to 825 m caused an increase
in PAH emission concentration (g/in^), emission rate (g/hr), and era
factor (g/kg of fuel burned).  Total WSDSS emissions did not show
altitude to be a major effect although the trend was in the same
direction.  Burnrate exerts an inverse influence on total emissions
(as seen in Figure 2); however, the statistical analysis did not show
this to be a major effect, probably because of the wide variability.
The statistical analysis also confirmed the opposite, direct trend of
PAH emissions versus burnrate but not as a major effect.  Another
major effect was the direct relationship between stack flow rate
(normal m-Vhr) and burnrate.  On the other hand, increasing altitude
seemed to result in reducing stack flow rate.

Conclusions

     In the IACP field studies, emission tests on residential sources
such as woodstoves are necessary.  However, some variables, such as
burnrate, are nearly impossible to measure over short time frames of a
few hours without causing a major disruption to the residents.  The    .
parallel testing of such residential combustion sources under controll6
conditions in a laboratory offers the advantage of allowing measurement
of all parameters under simulated field conditions.  Together, the
field and laboratory data provide the means of adequately character! z*11*
these sources.

     Combustion in woodstoves is an inherently variable process becausfi
of the nonhomogeneity of the fuel and the batch nature of the fueling
procedure.  Including cold start in the test protocol adds substantial
more variability.  Even with proper statistical test program design*
this variability makes drawing conclusions difficult.

     This project showed that PAH emissions from a woodstove typical °
those used in Boise, ID, were higher at Boise's elevation than at neflf
sea level.  When completed, the bioassay results may shed more light °n
these findings.

References

1. 40CFR Part 60, Standards of Performance for New Stationary Sources*

                                   838
-------
    Standards of Performance for New  Sources,  Residential Wood  Heaters;
    Federal Register,  February 26,  1988,  pages 5860-5926.

 2.  Oregon Administrative Rules, Chapter  340,  Division 21,  -100 through
    ~~ i y u *


 3.  Colorado Air Quality  Control Commission  Regulation 4, Regulation on
    the Sale of New  Wood  Stoves, June 27,  1985.

 4.  Leese, K.E. and  R.C.  McCrillis, "Integrated  Air Cancer  Project  -
    Source Measurement,"  in Proceedings:   79th Annual Meeting,  APCA, Paper
    No. 86-74.7, Minneapolis,  June  1986.

 5.  Merrill, R.G. and  D.B.  Harris,  "Field  and  Laboratory Evaluation of  a
    Woodstove Dilution  Sampling System," in  Proceedings:   80th  Annual
    Meeting, APCA, Paper  87-64.7,  New York,  June 1987.

 6.  40CFR Part  136, Appendix A, Method 610 - Polynuclear Aromatic
    Hydrocarbons.

 7.  McCrillis,  R.C. and R.G. Merrill, "Emission  Control  Effectiveness of
    a  Woodstove Catalyst  and Emission Measurement  Methods Comparison,"  in
    Proceedings:   78th Annual  Meeting, APCA, Paper No. 85-43.5  Detroit
    June  1985.
 ,
c
•^
01

$
a
L.
E
Ul
    60
    50
40 H
                            Nonextractable
                            Semivolatile
                            Condensible
     Test codes
first letter -
  S - conventional stove
  E • catalytic stove
second letter -
  0 - oak
  P • Boise pine
third letter -
  L - 90 m altitude
  H « 825 RI altitude
fourth letter -
  L • low burnrate
  H - nigh burnrate
      SOLL"1
       o™,  n
       SOLl-2
                 SOLH-2
                                    SPLH-l     SPKL-l     SPHH-l      EPHL-1
                               SPLL-2     SPLH-2     SPHL-3      SPHH-2     EPHH-1
                                  Test Code
     Figure 1.  Bpise source laboratory emission test results  showing  total
               dilution sampler emission nates for nonextractable.
               semivolatile.  and condensible fractions.
                                    839
-------
     60


     80 -


»   40 H

a"
5   30 H
S
8
     10 -
                 Circled No.  Total  mod stove dilution aanpler (NSDSS) train
                     *     Total  WSDSS condenalble orgsnica
                     +     Total  WSDSS aenlvolatlle organics
                  (Burn Nos. 12 fi 13 are catalytic stove.
                  All others are conventional stove.)              A
       l.l
              1.3     1.5     1.7     1.9    2.1     2.3     2.5    2.7     2.9
                                   Burnrate. dry  kg/hr
                                                                                3.1
      Figure 2.  Boise  source  laboratory  emission test  results showing  the
                  effect  of burnrate  on emissions.
o
•r*

ID
£   i,«H
    0.5-
               Suti of 16 PAHs
                 hthalene
                 ene
                 nzo-a-Pypene
          (Solid symbols  nark results from catalytic  stove.
          All other results are from conventional stove.)
         --.-**-	
       1.1
1.3
                                     1.9     2.1     2.3
                                     Burnrate. kg/hr
                             2.5
2.7
                                    2.9
                                                                                3.1
     Figure  3. Boise source  laboratory emission  test  results  showing the
                effect of  burnrate on emission rates for  selected PAHs.
   0.04-
          (Solid symbols nark results from catalytic atove.
         All other results an* from conventional stove.)
                                                                    Pyrene
                                                                    Benzo-a-Pyrene
        1.1
               1.3     1.5
1.7
1.9     2.1    2.3     2.5
 Burnrate.  kg/hr
2.7    2.9
    Figure 4.  Boise  source  laboratory  emission test results showing
                rates  for pyrene and  benzo-a-pyrene versus burnrate.
                                          840
-------
      OF RESIDENTIAL WOOD
 JJIBUSTION AND AUTOMOTIVE EMISSIONS
   THE BOISE, IDAHO, AIRSHED

 ^ Ross Highsmith, EMSL
 jj B. Zweidinger, ASRL
 U e4-len Lewtas , HERL
 He S* EPA
  search Triangle Park, NC 27711
       wisbith
   Associates, Inc.
  99 Cheater Road
  cinnati, OH
       J.  Hardy
?0a  S0n-Knudsen Engineers
I* park Blvd.
 i8e»  ID 83709


fj.0  large-scale ambient  monitoring program was conducted in Boise, Idaho,
    Ausust !986  through February 1987 to evaluate the impact of residential
 c  Combustion (RWC)  and  automotive emissions  on the local  airshed.   Con-
 f "-Ve 12 h  samples  were collected at three primary sampling  sites:   (l)
       ntial neighborhood impacted  by  RWC  emissions,  (2)  the vicinity  of a
       traveled  Boise  intersection, and  (3)  a desert  area outside  the
    airshed.   Particulate,  semi volatile  organic,  and  volatile  organic
 Hd     were collected at  each primary  site.   Particulate  sampling was  also
        at four auxiliary sites located throughout the city.   The  auxili-
  8*te  data  were  collected  to assist  in the  overall  evaluation of  the
       8ites as  well aa "to provide  information  regarding  the uniformity of
     S*  Routi-ne  criteria pollutant  and meteorological parameters were
   "tonitored.   The samples have  been  analyzed for particle  mass,  metals,
 0 Otl» organics, and mutagenicity.  Comparisons of key pollutant concentra-
    ^e*ween monitoring  locations  have  been  conducted.   Background fine
       concentrations  average  10  ng/m3.    Fine particle   concentrations
       100 ug/m3  during  several winter  nighttime sampling periods when
  Missions were increased.  Fine particle lead, coarse particle,  and  NOX
  ®ntrations were elevated at  the mobile source impacted  primary site.  An
-ft      of the monitoring program as well as a summary  of key findings is
  ented.
                                  841
-------
               IMPACT OF  RESIDENTIAL  WOOD  COMBUSTION  AND  AUTOMOTIVE
                     EMISSIONS  ON  THE BOISE,  IDAHO,  AIRSHED

Introduction

     The influence  of mobile source and  residential wood  combustion  (°
emissions on  ambient  air quality  has  been previously reported. 1-b
atory studies**-** have  demonstrated that both  sources  generate high
centrations of  fine particles  (aerodynamic  diameter  <  2.5  (am)  that  are
in polycyclic  organic matter  and products  of  incomplete combustion.   ^
occurrence and contribution of each source can be  readily  distinguished 1
other sources and  is  easily correlated with the  concentration of  a  un  >0
inorganic tracer.   Fine particle lead  concentrations  have been tracked
show that  mobile  source  emissions  are  highest  during  weekday   samp1
periods with maximum  emissions occurring  during  peak  rush hour per
RWC emissions,  rich in  fine particle potassium,  typically maximize
winter nighttime sampling periods.

     During the 1985-86 winter heating season the  Integrated Air Cancer
ject (lACP)itJ conducted  a  series of pilot  studies in predominately
source impacted airsheds.   The objective  of these pilot studies was  to  ^
velop monitoring and analytical protocols  for identifying  and character*2 ^
the impact of carcinogens on both indoor and ambient air.  State-of-t*16' $
monitors for particulate,   organic,  and gaseous  pollutants  were evaiua
and operated to collect data in support of the I AGP objectives. The
of these pilot studies supported the use of the IACP procedures  to char
ize airsheds impacted by multiple sources.   A large-scale monitoring Pr  
lation approximately  150,000.    It  is  the  governmental,  educational'^s'
commercial distribution center  for  the state and  region.   No major   , tf
trial sources  are  located  in  or  around  Boise.   The  city (Figure    t"
situated along the  Boise  River  basin, which  traverses from  south6*  J-F
northwest, and is  bordered distantly  by  the  Rocky Mountains  to  the   -jt'
and east and by a  series of  elevated benches (plateaus),  each rising a^

                                   842
-------
4U H^ ^°.m>  to the west*  Desert  area immediately surrounds the city  in
ret   rections*   The topography  and  meteorological conditions  favor the
  ention of local emissions,  especially during intense winter inversions.
  Ore detailed description of Boise is  provided elsewhere.12

    TW° Primarv  fixed-site monitoring  stations  were  established during
    * 1?86'   °ne WaS  located at  Elm  Grove  Park (EGP), a  central bench
      tial area imPacted by RWC emissions  during the winter heating season.
   second primary monitoring station  was  set up at a background site, the
       Aviation  Administration's  Radio-Controlled  Air-to-Ground  (RCAG)
     •  '  located in  the desert  approximately  7 miles  south  of  the city.
    fixed-site  monitoring  station  was  established  in  accordance  with
         Particulate Network (IPH)13 and MAMS/SLAMS11* criteria.  The types
Htt nu«ibers  of  samplers operated at each  site are shown in Table I.  Consec-
c0ftde 12 n ambient sampling periods (changeover at  7 A.M.  and 7 P.M.)  were
V6l.g  ted  from  August 10-26 and  September 7-15, 1986,  to  evaluate  daytime
^ US ^Shttime source contributions.  Sampling procedures developed during
*tfce   ^ IACP studies were  employed  to minimize sampler downtime  and maxi-
t>H   sampling time.   Samples for mass and inorganic analysis  were  collected
^a   Veighed 1uartz  and Teflon®  media.    Samples  for bioassay and  carbon
       were collected  on Pallflex T60A20 Teflon®  impregnated  glass fiber
     and Pallflex QAOT quartz   filter  media, respectively.   Vapor-phase
            organic compounds  (SVOCs) were collected using XAD-2 absorbent-
     canisters installed immediately downstream of the  particulate  filter.
         impregnated with 2,U-dinitrophenylhydrazine (DNPH)  and  evacuated
     polished  canisters were used  to collect aldehydes and volatile or-
  j° Compounds  (VOCs),   respectively.  Prototype annular denuder sampling
  ns   were used  to collect  acid aerosols.   Routine  criteria  pollutant
  "^teorological  parameters  were  also  monitored  according  to  reference
           SVOC, bioassay, and carbon samples were stored at -80°C iraraedi-
           3amPlinS-  Samples  for mass determination were conditioned for
      °C, kO% relative  humidity) and  then processed by standard  gravimet-
      ysis.   Aldehyde  samples were  capped and stored at -1*°C.   Annular
  .er_sample trains were disassembled and  the tubes immediately extracted.
   ^dividual denuder filters and extracts were stored separately at -tt*C.

     third mobile  source impacted primary monitoring  station was  estab-
     ^Or  ^ne  winter  study in November  1986  on the  second-story roof of
  t   tion #5  (FS)  located at  the corner of Front  and l6th Streets.  Week-
   ^affic  on  both  commercial  corridors  is  reported to  approach  10,000
   T"68 per  day.  Four  auxiliary  monitoring  stations,  located throughout
     Sned>  were also  established in November 1986.   The auxiliary site data
    °Hected to provide information regarding the  uniformity of emissions
      presence and direction of any regional  source contributions  during
         monitoring study.   Two of the auxiliary stations (Camelback Park
      s School) were located in central bench residential neighborhoods.
      3tation> .Winstead Park, was established in  a northwest  residential
       ood  located  on the first bench.  The fourth auxiliary  station  was
      at  tne Western Idaho  Fairgrounds, a northwest area with light-to-
  8i  daytime commercial activity.  The  Fairgrounds site was  bordered on
    des  by  heavily  traveled streets  (daily  traffic count  approximately
   >ellicles  each street).  Each fixed site was established  in accordance
    ihalable Particulate Network  (1PN)13 and  NAMS/SLAMS14 criteria.   The
  catld numbers of samplers operated  at each site are shown  in  Table I.
    nfiguration of samplers at  the primary monitoring sites was identical.
         12 h ambient   sampling  (changeover  at 7 A.M. and  7 P.M.)  was
       at the tnree  Primary and  four auxiliary monitoring stations from
      3, 1986, through February  6, 1987.  The sampling procedures  and
      media employed during the winter study were  identical to those used

                                  843
-------
                                                                        aj
 during the summer  study.   A more  detailed  description  of the  sites
 sampling protocol is provided elsewhere.16*17

 Analysis

      The analytical procedures used in this study are referenced in Table  ^
 Eighty EGP and 80 FS samples were selected for detailed analysis.  These V
 samples, collected over consecutive  Wday weekend/ weekday sampling P®ri ^
,(7  A.M.  on  Saturday through  7  A.M.  on  Wednesday),  correspond to  t&e
 weeks of the  winter residential  sampling study.11   This subset included
 weekend daytime,  20 weekend nighttime,  20 weekday daytime,  and 20
 nighttime samples for each primary site.  For each site, the paired
 daytime,  weekend nighttime, weekday daytime, and  weekday nighttime me •  (
 flow (0.113 m3/min) particle  and XAD-2 SVOC  samples were pooled,  and,.ii'
 pooled samples independently solvent-extracted with  dichloromethane.
 quota  of each extract were evaporated to dryness for gravimetric determ*,
 tion.   Additional aliquots were solvent exchanged with dimethylsulfoxide  ^
 stored at -80°C in  preparation for bioassay.   The Ames bioassay,18»i9   ^
 and without the activation agent 39, was used to determine sample mutag  ^
 activity.  Gravimetric, aldehyde, and VOC determinations  were  conducted  fl
 individual samples.  Where appropriate,  these results have been  pool6
 facilitate comparisons with the SVOC and mutagenicity results.

 Results and Discussion

     Nearly 1*0, 12  h sampling runs were  conducted during  the summer  st  .,
 Table  II summarizes the resulting particle concentrations and  size A 8   '
 butions.  The close proximity of  brushfires  during the initial 2 wi
 the study necessitated a temporary  shutdown  of the  project.  The  mass
 centrations observed during the period impacted by brushfires approxin1*  ,t
 doubles the concentrations observed during the period without brushf1j,
 For both periods  the data  reflect small differences in particle  size dis  ^
 butions or loadings between the  two primary   sites.   Little variabil^^f
 seen between  daytime  and  nighttime  periods.   The  large  coarse-*0' gf
 particle ratio is typical  for arid environments impacted  by windblown »  ,,»
 This ratio decreased slightly when the brushfires were present  (1.65  ^e $t
 1.90), suggesting an  increase  in the  fine particles associated  witb ^
 brushfires.  Of more importance,  the PM^Q-to-TSP ratio for both periods  (
 0.55*  indicating the aerosol  source remained  relatively constant throug  ,(
 the summer study.   The large  coarse-to-fine  ratio, the high TSP  load3- „&
 and the uniformity   in particle size distributions between the  EGP am
 sites  suggest that  the desert was  the dominant source impacting  the
 airshed during this phase.  The summer study data suggest that the bacfc
 fine particle concentration for the Boise airshed is approximately 10 ^

     Average daytime and nighttime particle concentrations observed  *
 three  primary and four auxiliary  sampling  locations  over the more than(.0l»/
 winter study sampling periods are provided in Table III.  Fine particl6
 centrations in excess of  100 ug/m^  were observed in  the  night durinS
 Fairgrounds and three  nighttime  EGP  sampling periods.   The  highes*
 particle concentration observed  was  li*0 ^g/m3  at  the   Fairgrounds
 These  levels are  significantly  lower  than the >200  ug/m3  fine
 concentrations observed during  previous  winter heating seasons  by  the * iy
 agency.  12 The lower than  expected fine  particle concentrations are ** jj'
 uted to an extremely mild Boise  winter  and to the newly  passed cit/ ^f
 nances limiting woodburning during periods  of intense meteorological   ^
 sion and high  particle concentrations.  EGP  fine particle  concentr*
 correlated with the FS concentrations.  Nighttime fine particle  cor
 tions  were nearly  50$  larger than  daytime concentrations at  both

                                   844
-------
 ^se-to-fine ratios observed during this period (ranging from 0.2 to 0.7)
. re less than half  the summer study values.   This  shift in particle size
 atribution is  attributed to  three  factors:    increased  RWC  emissions,
 Creased desert  contributions,  and  decreased  mixing  that  resulted from
'  utime temperature inversions.  Elevated coarse particle and TSP concen-
        indicative  of  automotive  traffic  were  observed at  the FS  site
8,
  s
 K lng daytime  sampling periods.   Daytime and nighttime RCAG  fine  particle
  vels were  nearly identical and approximate the summer background  values.
  e RCAG  coarse particle loadings decreased  by  a factor of 3-5  during  the
   er.   Overall, the RCAG data suggest the absence of any regional source(s)
 y 'Ughout the  winter  study.   The  auxiliary site data  indicate that  the
 . rshed was  uniformly  impacted by the local emissions.  The two northwest
     ary sites  were  slightly more impacted by RWC and mobile  source emis-
     and typically reported the highest fine particle loadings.   This find-
 ,j8 compares  well with previous  local agency findings.12  The Fairgrounds
   > located in close proximity to  heavily traveled  roadways, was  similar
      FS both  in particle loadings and in  the  distribution  of particles
       the fine and coarse modes.  Daytime Fairgrounds samples were  heavily
    enced by large coarse  particle  concentrations which are attributed to
         mobile source activity.

    The results of inorganic  analyses on selected primary residential and
      source sites samples are shown in Table IV.  These data reflect only
 Vy °0 sampling periods selected  for  detailed analysis  with  each average
 KQ Ue representing 20 individual values.  Extended analyses planned for the
 m^* and auxiliary  site samples have not been completed.   The average EGP
 OK  ime fine  Particle potassium  (corrected for  soil contribution)  was
 .'l& "rr/jjjS and  is  nearly  twice the daytime  value.  The increased potassium
      correspond with increased RWC emissions.  The EGP fine particle potas-
    levels  were consistently higher (0.03 (ig/m3) than the corresponding FS
    ntrations.   The fine  particle mass  and fine particle  potassium data
   ected from both primary sites were independently  subjected  to linear
      ion "best-fit"   routines.   Both  curves  yielded  nearly  identical
      (0.0035)  and correlation coefficients exceeding 0.91-   Comparisons
  n8  only the nighttime samples improved these correlation coefficients  to
 !0'^«   Similar  comparisons  for fine particle mass and fine particle  lead
  Cfintratlons were less precise  (correlation coefficients  <0.75)*   The  FS
         slope (0.0013)  doubled the  EGP  value,  suggesting  that  the  FS
        was   impacted   by  mobile  source emissions.   Comparisons  of NOX
      fine mass for both primary sites yielded results nearly  identical  to
   linear regression  results  for  fine lead.  Again,  FS NOX  concentrations
   "  'mately doubled the  EGP values.  NO comprised approximately 60% of the
      concentration but only 50% of  the EGP NOX  concentration.  The higher
      NOX concentrations observed at the FS  site are attributed  to mobile
      emissions.   CO concentrations  correlated poorly (correlation  coeffi-
      <0.68)  between  the two  primary  sites.  Only minimal differences  in
iv^ged  CO  concentrations  were  observed  between the  two  sites.   Linear
          between fine particle mass  and  CO concentrations also  yielded
   correlations  (<0.70)  for both sites.

   Approximately  60%  of  the fine particles  collected  at  EGP  and FS were
          with  dichlororaethane.   This  result   compares  favorably with
      IACP results.   Ambient formaldehyde concentrations averaged  approx-
      3-lt ppb   for both primary  sites and  did  not  vary  between   sample
 „a   or  site.   Total  carbonyls averaged 10-20 ppb and were not dependent
 jt^P^ng period.  The FS total carbonyl averaged values were consistently
    higher than the corresponding EGP  samples.  Weekday FS VOC concentra-
 Vn averased  803 ppb C, nearly  twice the weekday EGP values.   Weeknight
   C concentrations  were  more  than $0%  higher  than  EGP  VOC  levels.

                                  845
-------
 Comparable daytime and nighttime  weekend  VOC concentrations (approximate
 ^25  ppb C) were observed  at both primary sites.  The  increased FS weeKdw
 values  are most probably  associated  with the mobile  source contributi-011 '
 The  particle sample mutagenicity ranged from 60 to 85 revertants/tn^ wit*1
 differences observed between  sites.   A slightly  higher mutagenic act^
 (10  revertants/rn3)  was observed  daring  weekday periods at the  FS site a
 is attributed to local automotive emissions.  The mutagenicity of the
 extracts  was highly  variable and  was  biased  by  a  large blank

 Conclusions
                                                                        * $
      The  Boise background fine particle concentration  remains  constate
 approximately 10 ^g/m3.   Samples  collected at the two summer study sampl
 locations were nearly identical  in both particle concentration and part ic
 size distribution.   Both summer study sites  were heavily impacted by ^ n5
 blown desert dust with little or no additional  regional source contribu^
 observed.
                                                                         tltf
      Mobile source  and RWC emissions  were  the primary sources impacting^  -
 Boise airshed during the  winter  heating  season.  Particle  concentrate
 were uniformly distributed across the city.   Slightly higher  concentrate
 were observed at the northwest sampling locations.  Nighttime fine
 concentrations were nearly $0% higher  than daytime values at all the
 and  auxiliary sites.  This is attributed to increased RWC emissions and
 occurrence  of inversions during nighttime  sampling periods.   Fine par*1 j
 concentrations in excess  of  100 pig/m3 were observed in both residential e
 commercial  areas during winter meteorological inversions.  Elevated EGP  j.y
 particle  potassium   concentrations  suggest  that  this  residential  Pr* n)
 sampling  site was impacted by RWC emissions.   Excellent correlations *   '$
 between fine particle mass and fine particle potassium were observed at  ,$
 EGP  and FS,  suggesting that  RWC impacts  the entire airshed.  Increased * ?
 lead,  coarse particle, and NOX concentrations  were observed  at  the FS  s
 indicating  the presence  of  mobile  source  emissions.  Extractable
 comprised nearly 60% of the fine particles  collected at both the re
 and  mobile  source sites.   Total carbonyls  and total VOCs were higher at ^
 FS site during weekday and weeknight sampling periods.  Formaldehyde **&&*»
 trations  did not appear to be influenced by  either sampling period or 3 -^
 ling  location.   The background site  data  indicate minimal regional  3°U
 contributions present during the  winter  study.
Acknowledgement

     The authors  thank  Ralph  Baumgardner,  Atmospheric  Sciences  ReS  et<
Laboratory /EPA for  coordinating   the   inorganic   analyses;   Linda  ^Ol0j&t*
Morrison-Knudsen Engineers,  for coordinating  field quality control »n
-------
 lt   V. R. Highsraith, C. E. Rodes, R. B. Zweidinger, R. C. Merrill, "The
     collection of  neighborhood  air  samples  impacted by  residential wood
     combustion in Raleigh, NC and Albuquerque, NM," 1987 EPA/APCA Symposium
     on Measurement of Toxic and Related Air Pollutants, Air Pollution Con-
     trol Association, Pittsburgh, PA, 1987, pp. 562-572.
 g
     K. Sexton, J. D. Spengler, R. D. Treitman, W. A. Turner, "Effects of
     residential wood combustion  on  indoor air quality:   A case  study in
     Waterbury, Vermont, Atmos.  Environ. 18 : 1371-1383 (1984).

     D. J. Moschandreas , J. Zabransky, H. E. Rector, "The effects of wood-
     burning on the indoor  residential  air  quality," Environ.  Int. h: k&3~
     ^68 (1980).
 4
     J. N. Pitta,  J. A.  Sweetman,  W. Harger, D. R. Fitz,  P. Hanns-R, A.  M.
     Winer, "Diurnal mutagenicity  of airborne particulate organic  matter
     adjacent to a heavily  traveled  West Los Angeles Freeway,"  JAPCA  35_:
     638-6U3 (1985).

     R. K. Stevens,  C. W.  Lewis, T.  G. Dzubay,  R.  B.  Baumgardner, L.  T.
     Cupitt, V.  R. Highsmith, J. Lewtas, L. D. Claxton, B.  Zak,  L.  Currie,
     "Source apportionment  of mutagenic activity of fine particles collected
     in Raleigh, NC,  and  Albuquerque,  NM."   Presented  at  1987 EPA/APCA
     .Symposium on  Measurement of Toxic and  Related Air Pollutants,  Air
     Pollution Control Association,  Pittsburgh, PA, 1987.

     D.  G.  DeAngelis, D.  S.  Ruff in,  R. B. Reznik,  "Preliminary character-
     ization of  emissions from wood-fired residential combustion equipment,"
     EPA-6(X)/7-80-OUO.   U.   S.  Environmental  Protection  Agency,  Research
     Triangle Park,  NC.   Available as PB 80-182066 from National Technical
     Information Service, Springfield, VA, 1980.

     B.  R.  Hubble, J. R. Stetter, E.  Gebert, J.B.L. Harkness, R.  D. Flotard,
     'Experimental measurements  of emissions  from residential  woodburning
     stoves,"  Residential  Fuels; Environmental  Impacts and  Solutions
     (edited  by J. A. Cooper, D.  MalekJ.  Oregon Graduate Center,  Beaverton,
     OR, pp  79-138 (1982).
 8.
     s* S. Butcher, E. M. Sorenson, "A study of wood stove particulate
     emissions,"  JAPCA 19:  72U-723 (1979)*
 9,
     c- W. Lewis, W. Einfield, "Origins of carbonaceous aerosol in Denver
     a1d Albuquerque  during  winter,"  Environ. Int.  11:   2l* 3-2^7  (1985)-
l0'    ,
     J. Lewtas, L. T. Cupitt, "The Integrated Air Cancer Project:   Program
    overview," 1987 EPA/APCA Symposium on Measurement of Toxic  and Related
                  , Air Pollution Control Association, Pittsburgh,  PA, 1987,
    v- R. Highsmith, C. E. Rodes, R. B. Zweidinger, J. Lewtas, A. Wisbith,
    **• J.  Hardy, "Influence of  residential  wood  combustion on indoor air
    Duality of  Boise,  Idaho,   residences,"   1988 EPA/APCA Symposium on
    jfegjurement of Toxic and Related Air Pollutants, (In press) Air Pollu-
    tion Control Association, Pittsburgh, PA, 1988.
                                   847
-------
 12.  L. T. Cupitt, T. Fitz-Simons, "IACP Boise field program:  Study
      and survey  results," 1988 EPA/APCA Symposium on Measurement of
      and Related Air Pollutants, (in press)Air  Pollution Control
      tion, Pittsburgh, PA, 1988.

 13.  U.S.  Environmental Protection  Agency,  Inhalable Particulate Nettfg£x-
      Operations and Quality Assurance Manual, Environmental Monitoring   ^
      Systems Laboratory,  U.S.   Environmental  Protection  Agency,  Reseal>
      Triangle Park, NC (March 1983).

 lU.  U.S. Environmental Protection  Agency,  Federal Register, Vol. _U£ W
      10^08 (March 10, 198U).

 15.  R. K. Stevens,  T.  G. Dzubay, R. Baumgardner,  R.  Zweidinger, R. ^ .
      smith, G.  Fortune,  W.  Ellenson,  R. Hardy,  "Annular  denuder
      from Boise," 1988 EPA/AFCA Symposium on Measurement of Toxic and
      Related Air Pollutants,  (In press) Air  Pollution  Control
      Pittsburgh,  PA, 1988.
 16.   PEI Associates, Inc.  "Integrated Air  Cancer  Project  Summer Wood31"  {
      Study,  Boise,  Idaho," Report  under  EPA  Contract  68-02-^195  (Oct°
      1986).
 17-   PEI  Associates, Inc., "Integrated Air Cancer  Project  Winter
      Study,  Boise,  Idaho," Report  under EPA  Contract 68-02-1+195  (Oct°Dt"
      1986).

 18.   B.  N.  Ames, J. McCann, E. Yamasaki, "Method for  detecting carcin0^
      and  mufcagens with the salmonella/mammalian-microsome rautagenicity
      test,"  Mutat.  Res.  31: 3^7-361+ (1975).

 19.   D. M. Maron, B.  N.  Ames,  "Revised methods  for  the salmonella
      city test." Mutat.  Res.  113;  173-215  (1983).
20.   D. E. Lentzen,  et al,  IERL-RTP Procedures  Manual:   Level 1
      tal Assessment, 2nd  edTj  EPA-600/7-78-201.  Available  as
      (10/78) from  National  Technical Information Service, Springfield'
      1978, pp
21.  L. D.  Johnson,  R.  E.  Luce, R. G. Merrill, "A  spot  test  for p
     aromatic hydrocarbons,"  U.S.  Environmental  Protection Agency,
     trial  Engineering  Research  Laboratory,  Research  Triangle  pa**'
     27711.
22.  J. A.  Cooper,  L.  A.  Currie, G. A. Klouda, "Assessment  of
     carbon combustion  sources to urban air  particulate levels
     bon-lU measurements,"  Environ. Sci. and Techno!*_ lg_; 10^5-1050

23.  L. A.  Currie,  J. A.  Noakes,  D.  N.  Breiter, Proc. Int.  Conjj
     carbon Dating, 9th, 1979, p.158.


2k.  P. J.  Groblicki,  S. H.  Cadle, C. C. Ang,  P. A.  Mulawa,  "IntefljJ^
     atory comparison  of  methods for the analysis of  organic and eie& $
     carbon in  atmospheric  particulate  matter,"   General Motors  %e6  W
     Report No.  kQ^k.   General  Motors  Research  Laboratories,
     1*8090 (1983).


                                   848
-------
    R. W.  Shaw,  R.  K.  Stevens,  J.  Bowermaster,  J.  Tesch,  E.  Tew,  "Measure-
    ment of  atmosphere nitrate  and nitric  acid:   The denuder difference
    experiment," Atmos. Environ. 16: 8^5-853 (l98l).

    P.  D.  Stump,  D.  L. Dropkin,   "A  gas  chromatographic  method for the
    quantitative spec iat ion of C2-C13 hydrocarbons in roadway vehicle emis-
    sions," Anal^Chem^ 5J_: 2629-2631*  (1985).
'•  S. B. Tejada, "Evaluation of silica gel cartridges coated  in-situ with
   acidified, 2,^-dinitrophenylhydrazine for  sampling aldehydes and ketones
   in air,"  Int. J. Environ. Anal. Chem. 26: 167  (1986).
                                 849
-------
       TABLE I.   NUMBER OF AMBIENT SAMPLERS AT EACH FIXED-SITE LOCATION
ANALYSIS
MASS

BIOASSAY
SEMI-VOLATILE
ORGANICS
CARBON 14
c/c.
ACID AEROSOLS
VOLATILE
ORGANICS
ALDEHYDES
CONTINUOUS
™ ™nr- PRIMARY AUXILIARY ANALYSU
SAMPLER TYPE m S|TE REFB?0g
Dichotomous
^Pftn
TSP
PM2.6 Hi-Vol
PMID Medium Flow •
PM|0 Medium Flow
with XAD-2 •
PM2! HI-Vol
Modified Dichotomous •
Annular Denuder •
Evacuated
Canister
DNPH
CO, NO,, 03, BKlt
WS, WD, T, RH, SR
2 1
'
4 1
1

2
1
1
1

2
YES WS and WD
YES only
13

18, 19
20, 21

22, 23
24
15,25
26

27
^t*
  Not operated at RCAG
TAbLC. 11. OUMMrin bTUlJI AiVimiWr VJUHUCiNrnATlUno S
SITE
ELM GROVE a

RCAG a


ELM GROVE b

RCAG b

SAMPLE
PERIOD
DAY
NIGHT
DAY
NIGHT

DAY
NIGHT
DAY
NIGHT
FINE
8.4
8.4
6.7
6.4

18.8
13.8
15.7
13.4
COARSE
15.2
14.3
12.9
14.3

26.1
27.6
22.6
25.2
TSP}
44.^
32-^
46^
37-^
__ ^~~^
Q^t£ .
Q^y
l*t/
1Q$/
"   Without brushfires
b   With brushfires
                                   850
-------
          TABLE III.   WINTER STUDY AMBIENT CONCENTRATIONS
SITE
,^LM GROVE
• 	
^IRE^STATION
--—-__
^CAG
:==^
^DAMS SCHOOL
••— —
^AMELBACK PARK
	
^IRGROUNDS

.WJNSTEAD PARK
* 	
SAMPLE
PERIOD
DAY
NIGHT
DAY
NIGHT
DAY
NIGHT
DAY
NIGHT
DAY
NIGHT
DAY
NIGHT
DAY
NIGHT
FINE
/ug/m3
26.7
42.7
24.2
38.6
10.9
13.8
24.2
39.9
21.3
28.5
26.1
48.9
28.1
44.9
COARSE
jug/m3
7.9
6.2
16.3
11.1
5.5
3.6
8.2
6.3
8.6
6.1
13.0
9.8
9.6
8.5
COARSE TO
FINE RATIO
0.30
0.15
0.67
0.29
0.51
0.26
0.34
0.16
0.40
0.21
0.50
0.20
0.34
0.19
TSP
Aig/m3
52.8
63.3
82.5
77.9
26.0
21.6
NM
NM
NM
NM
NM
NM
NM
NM
'indicates not measured
 TABLE iv.   WINTER  STUDY  AMBIENT INORGANIC CONCENTRATIONS
PERIOD

WEEKEND
DAY


WEEKDAY


WEEKEND
NIGHT


WEEKDAY
NIGHT

CONSTITUENT
FINE MASS
COARSE MASS
FINE K
FINE Pb
FINE MASS
COARSE MASS
FINE K
FINE Pb
FINE MASS
COARSE MASS
FINE K
FINE Pb
FINE MASS
COARSE MASS
FINE K
FINE Pb
ELM GROVE PARK
25.4
6.5
0.09
0.04
27.1
9.4
0.10
0.05
43.4
5.9
0.15
0.04
36.4
7.1
0.16
0.04
FIRE STATION
21.2
9.5
0.06
0.05
26.4
22.8
0.06
0.10
36.6
9.4
0.12
0.07
37.9
13.3
0.13
0.07
           Etch ivcrig* vilui r«pr«Mnti 20 Individual vtluti
           from only Iht 80 utoctvd timpllng period*.
                                     851
-------
           FAIRGROUNDS
             EL. 2625 ft
                          GROVE PARK
                          EL. 2690 ft
CAMELBACK PARK 7
    EL. 2950 ft
                       TWWTlfflB
                   WINSTEAD  PARK
                      EL. 2700 ft
                                                    ADAMS
                                              ELEMENTARY SCHOOL
                                                    EL. 2700 ft
    BIRDS OP PREY
                                                   SCALE       ^
                                                 0       1 mil* A
                           FAA-RCAG SITE
                              EL. 3140 ft
Figure 1.  Map of  Boise showing Winter WoodsinoK.e otudy
                              852
-------
     SHOULD  WE  MEASURE?

  °sol Data: past and future
U4*-  Currie,  K.  R.  Beebe, and G, A. Klouda
  lonal Bureau of Standards, Gaithersburg, MD 20899
                                ABSTRACT

      Sampling designs  and  measurement  designs  are  fundamental  to  all
       measurement programs.   By definition they must  constitute the first
^a     To   illustrate   the  value  and  assumptions  involved  in  planned
»etQUrement,  we shall treat a very important,  practical problem  related to
*N>1°*  source  apportionment.    The  basic  task  was  to  optimally  select
^tic 6s ^or   c measurement from the  Boise  field  study of  the  Integrated Air
V * Project  (IACP).  The decision  making process was not  trivial,  in that
Ju$t e&ning  of "optimum"  in  this context had  to be considered,  as well  as
S^MfW^at   should  be  optimized.     Also,   the  question  had   practical
Hft^  *-cance because  of the need to balance  information gained  with resource
*&r0 afcions.    In deciding which of the  three  hundred  odd Boise wintertime
»Ki °1 samples to  analyze, we were guided by the  following factors:  a)  the
      e --  validation of an  elemental tracer model for aerosol carbon and
        from  motor vehicles  and residential  woodburning;  b)  prediction
      s and  regression coefficient  standard errors as  influenced  by  the
   6  selection  design;  c)  physically meaningful null  and alternative
    6ses,  with  special  attention to  alternative functional  relations  and
  °U x°de^s*   Another critically important factor was  the knowledge gained
  [p,*" the  prior  IACP  ^C  measurements  on Albuquerque  -  Raleigh aerosol
                                  853
-------
          1.   C-14  SAMPLE SELECTION --  OPTIMAL MODEL VALIDATION DESIGN

 1.1. Statement of the Problem: Objective

        A primary objective of -^C measurement in selected aerosol sampleS  s
 to test  the  validity  of the  elemental (K,  Pb) tracer model for carbonac^
 species.  If the soil-corrected potassium and lead contents of the particl
 truly  reflect  the  wood burning  and  motor vehicle  carbonaceous  emissi0*1^
 then they  should  yield  receptor  modeling  results consistent  with  the tfy
 carbon  (WC)  and fossil  carbon  (FC)  results from  the  direct  ^C tracer-
 If cost  were of no concern,  the best  set  to select  for  (^C)  measure**
 would be obvious :  one should assay the entire set of samples .  This w
 impracticable,  and  a poor  use  of resources,  however.    Therefore,  we  ""
 consider the  selection  of  an  optimal  subset,  to  adequately validate  {
 element  tracer  regression  model.    In  doing  so,   we  must  address

 fundamental questions, such as: a) What do we mean by "validation"?  b)  j?
 do we mean by  the  "optimal" subset, and what,  exactly,  is  to be optii"12^
 c)  How  do  we  determine  that  subset  (algorithm)?  and  d)  How does  s
 "validation power"  of the  optimal  subset vary with the number  of  sa^P
 selected?

                                                                          {
        Starting  point.     Before  considering  the  answers  to  the
 questions  in  the  next  theoretical  section,  let  us look  at  the
 point,  and  some external  constraints.    First,  as  shown   in  Fig.  *'
 discussed  in  section 2,  approximately 60-70 samples were collected from
 of  two  Boise   sites  [fire  station  (FS),  and  Elm Grove   Park  (EG)
 combination with each of two 12 hour periods  [0700-1900  (day), and I900'
 (night)].   Selection  criteria  given in  section 2 reduced these
 four sets of 30 for each  site-period,  to be  considered for  mu
 testing.   The final  four  sets  of samples, selected for  this purpose W
 EPA  were  somewhat   smaller,  consisting  of  about 20  samples  each.    j
 represented the  Initial  set for our  14C sample selection design.  (Note
 in Fig. 1, and  throughout  the text,  K refers  to  soil -corrected potassi^'
        Constraints.   Subset  selection  was  subject to three
equal  representation from the four site-periods, b)  adequate  carbon
for  meaningful  14C measurement,  and  c)  a  limit on  the  total nu
samples,   imposed  by  the  overall  "cost"   (time,  personnel,   anaiye ^
expenses,  etc.).    Constraint-b eliminated  a  few  samples  initially1,^'
constraint-c  set  a pro-tern  maximum  of  9  samples finally  from each s ^
period.   (This total of  36  samples does not correspond to  the  total ^ JC
of 14C  measurements, because  multiple chemical  fractions will  be measute (|i<
certain of the  samples.)   Two  circumstances  prevented  our  followi1^  ^l
classical  factorial design: first,  the  factor  levels  (K- ,  Pb-concentrac). f
could  not be  preselected;  second,  these  concentrations  could  not eVe
independently  selected, as they  were necessarily  coupled in  every s&W

1.2.  Some Theoretical  Issues

                                                                        <
        Here,  we   examine  a little  further the  meaning of  validati°°yf
optimal  sample subsets in the  context  of testing  the  two elemental C
model via  ^C.

        Validity  testing.   Validity  testing is,  In fact, hypothesis
so  we   must  consider,   in the  simplest case,  the nature  of the  nul
alternative  hypotheses,  and  the  operating  characteristic  of  ^e ^  .
applied.   That  is, given the  regression model  to be  validated  ^  '("
criterion  to be  used for  validity testing,  we can evaluate the
power)   of  the test in light of a  specified alternative  model.
tests and  multiple alternatives may be considered, but that  is bey0
current scope.)  For 14C sample selection, we began with the slmplest
                                   854
-------
          linear models  to  represent the  null  [H0]  and  alternative  [HA]


       H0:  WC - K bK  +  e    and   FC - Pb bpb  +  e                (1)

       HA:  WC - b0  +  K bK  +  e   and   FC - bo  + Pb bpb  +  e    (2)
 \t
 'tta i,.    if  mineral-corrected potassium  and  lead do,  in  fact,  linearly
 the,.*'  the  W°°d carbon    and  fossil  carbon (FC) aerosol  sources,  and
 Silrf^6 n° other  sources  of carbon,  then the 1-parameter equations  (1)
 ^ *•* correctly fit the 14C data,  which give direct measures for WC and FC.
 to  simplest alternatives,  Eq.  (2),  include intercepts (bQ),  or unaccounted-
 tftlat   °n (UFC) •   The  power of the test,  for UFC, can be shown to be simply
 fy  ed  to the SE of the estimated  intercept when the data are fit to Eq's.
 Oj ^  Other  HA's may also be of interest --  eg,  quadratic (in the variables)
 Cotisi0Jl"'Linear  ^ln  the Parameters)  models.    In any case,  unless an  HA is
 PtOof   ed|  consistency °f the data with  the null model cannot  be taken as
 86le   °f  validity-   These  facts  therefore directly  influence  the  sample
    ction  process.

      Ob1active function.   For a  given number of samples selected  for  14C
          '  we are concerned with  selecting the "best" subset.   "Best"  is
          defined  in terms  of an  "objective  function"  or  measure  which
  sni     a  maximum or  minimum for the  chosen  samples.   The  nature  and
 >9Udi  e  °f  thlS  function  can then  be  related to  the adequacy  of  the
 ^d Y.  y test-   We face  two  questions:  which objective  function to  employ,
 (1)   °w  to select  the best  sample  subset.    For  linear  models such as Eq's
 ^ ^nd  ^2)> a  natural  choice  of objective  function  is  the  determinant of
 f°ot fSi*>n matrix  |X| or,  for an overdetermined system,  |X'X| or its  square
        This  approach,  which  derives  from  sensitivity optimization  of
      raponent  methods  of  chemical analysis,  is  equivalent  to  the  "D-
        (Dopt)  criterion for  variance optimality  of statistical designs,
   o2ne  s®eks a maximum for  the   determinant  of the "Fisher information"
   mi °n the Parameter vector b.4  This is a good compromise, for  it yields
    lnimum volume confidence ellipsoid for the parameter estimators, but if
    s interested  in the optimal precision  for  a given parameter estimate,
      better  to use the  corresponding diagonal  element of  the variance-
          matrix<   Its  square root  is the  standard error for that estimated
        sucn as the intercept  in Eq.  (2).   (In some cases other optimality
        may  be appropriate,  such   as  G-optimality when one is  primarily
        with response  prediction intervals.4)

      The  following subsection ("Algorithm")  illustrates the  effect  of the
      °f obJective  function  on the specification of  the  optimal  samples.
        '  •^t ^s very important  to  recognize 3 key aspects of experimental
             TAE  ^timal  design   does   not  require  knowledge  of  the
            (14C)  results.   The design matrix X is  fully defined by the
    and the  values  of the independent  variables  (K, Pb).  (2)  The choice  of
      does not  affect  the validity of the  data  reduction procedure  once
  c S  Qre  obtained;  (3) A   design  always  exists;  if it  is  ad hoc  or
    gnized  it  is not likely to be very good.

tl|6    Ssjasitivlf-.y *ma1yciC|   Evaluation of  the  change  in the magnitude  of
**tpl  Active function  for the  optimal sample  subset with  the  number  of
    s (N)  comprising the  subset, constitutes a sensitivity analysis.  The
         P°wer wil1 increase monotonically for a given objective function,
        diminishing  returns.   At  some point  the marginal gain will not
    y the  incremental (generally fixed) cost, so it pays to go no  further
      nal COSt constraint (here, 9 samples for  each  site-period) may or
      set  in earlier.
                                   855
-------
1.3. Algorithm and Results.

                                                                       f tfte
       The algorithm  employed will  only  be sketched below,  because °l .
scope  of  this  paper  and  because  it  is  subject  to  continuing  resell ,
First, it  should  be realized that  a complete search for the best  subse*
N=9 samples  in a  full  set of, say,  20, is  a very large  task  --  given by ,?
binomial  coefficient.     (The  total number of combinations  C(20,9) e<*^\
167,960.)    The   method  employed   here  involved  an  approximately °P  y
                                                                       a
"seed," which  grew sequentially with  subset size  to  give an  approxWa
optimal subset for each N.  The first step of this  process  is  illustrate  j
Fig. 2,  which shows  the  variation of  the objective functions (  |X'X|i  (0
the 3  SE  multipliers) during  the  search for the best sample to be  adde
the N-2 seed, for the 3-parameter model,

            C - X b  - b0  +   K bK  +   Pb bpb  + e               (3)
                                                                        fO&
where  "C"  represents  combined (fossil  and wood  source) carbon, and  trie   .,
of matrix  X equal (1 K Pb) .   This  yields  the N=3  seed, and the search * ,
continues  to  identify the  best additional sample for N-4,  etc.  It  is c  flt
from  the  figure  that  different  samples  would  be selected  for diffe .$
objective  functions.   (The N=2 seed was derived  from recurring samples
the above  procedure was applied with replacement.)
                                                                       ,
       A small  complication was faced as we  sought the best samples  \o ^
N) for validity testing of Eq.  (1).   That  is,  K and Pb concentrations  ^
necessarily  coupled;   the   best  K- sample   subset  would  not  necessa ^
correspond to  the best Pb-subset.   One means of  compromise  is to b&se .
selection on Eq.  (3), which automatically treats the coupled concentrate ^
The  results  of  this strategy  are shown  in Fig.  3,  for  the  daytime  ^
Station (FSD) samples.  These  results were  derived using ((X'X)"1*)!!  '* $
                                                                       s* . b«
                                                                       ul
SE  multiplier  for b0  --  as  the  objective,  as  evidenced by  the
sensitivity function  of  SE(bo)  vs N.   Alternative "stopping rules"  c°l
applied,  such  as  total  cost,   marginal  improvement,  or  approach  c°  f)i
asymptote.   Here,  for  example,  more  than 75% of the  maximum reducti0  j
SE(bQ)  is achieved  by  the  time  5  out of  13 samples  have  been  seleC
provided  that optimal subset selection is  employed.
                                                                        A°f
       Final  sample   selection,  using   N-9   as  the  stopping   crite  Jfl
(equivalent to  a  fixed budget  rule),  was performed using D0pt, resulti^jw
the  sensitivity  functions  (for  FSD)  shown in  Fig.  4.   It  is  clear  ^
little is gained  by going beyond the optimal set  of 9 samples:  the fu*  $
(N-13, FSD)  would yield  further SE reductions of only  4%  for bo,  2-7* (j,e
bK,  and  <0.1%  for bp^.   For  comparison  with Fig.  1,  Fig.  5 depict^i  ..
distribution  of  all  selected  samples on  the  K-Pb  concentration  f  ^
Although  the  sample  selection strategy is  imperfect,  the formal  stt^   $
and  explicit design criteria  provide  considerable  improvement  ovef  M
subjective   approaches.      Research   underway,   involving   multicti   A
optimization  and  errors-in-vartables  considerations, will  make the V*°^
even  more  efficient in terms  of  information   gained  vs  funds   eXPe^n t>*
Finally,  it  is  important to recognize that  the  design  (Boise  samples
assayed   for  ^^C)  in  no  way   affects   the  validity  of  the
experimental results; it affects rather the potential information  cont*
those results.                            	
                  2.  CALIBRATION SAMPLE, OC.LE.UIJ.UIN STRATEGY

       As noted  above,  the   C sample selection algorithm operated  0°  . {,
-------
,Ssemblage   of  Fig.   I.     The  philosophy  (algorithm)   adopted  for  this
 Vibration"  design  differed  somewhat from  that applied  to  "validation,"
c ^e  we sought to  optimally  test  the null  model  [HQ]  (given HA) .   The
  '-oration selection algorithm is outlined below.

,,.      The first step of  the  selection process was the  removal of samples
 ^ a volatilizable carbon measurement of less than 3.0 /ig/m3  as  this  was
       ed   the  minimum  amount  of  material  required  to  yield  reliable
  agenicity results.   This  resulted  in  the  deletion  of a  total of  32
 t
t     One  of  the  criteria  used to  select  samples  was  that  the  subset
„ Present the  "normally expected"  levels of  K  and  Pb .    It  was  decided,
8, Before,  that  samples with  extreme values  of  either  of  these  elements
t,°uld be deleted.  (Samples with  large  values might  not  be expected to fit
v  linear mutagenicity model.  The  goal was  to determine the model for the
Val6 fcypical samples even if it was not valid for the  full possible range of
«4    • )    Frequency  histograms for  the concentrations  of  K  and Pb  were
Vsan»lned  and all  samples that exceeded the 97.5 percentile for concentration
-------
                                                                         bo*
       4) Next,  one  or  two additional replicates near the center  of  trie
were included.  [4's  in Fig. 6d]                                          ^
       5) The design was completed by adding points  from  the  interior ot  ,
pseudo-factorial design.   These were chosen so  that  a good representation
all of the observed  samples would be present.  [5's in Fig 6d]

       In  total,  25 design  points were  chosen for each site and samp
period.    In addition,  5  optional   samples  were  selected  to be  use »
instances where a design sample was unavailable.  When an optional poinc
required,  we recommended  use  of  the  most similar optional  sample C '
minimum of the  sum  of  squared differences of concentrations  of K  and ^
replace the  rejected design point.   As in Fig. 3a,  the sensitivity funC ,*«
in Fig. 6c lacks smoothness,  reflecting the fact that the calibration »e
strategy was not optimal, in the sense of maximizing Det(X'X)

             3.  INFORMATION  DERIVED FROM THE PAST:   ALBUQUERQUE SAMPLES

3.1. Measurement Error
                                                                      ,  tfift>
       By  examining  the  observed  measurement imprecision  associate^  . jj
prior  estimates of wood carbon (WC) and fossil carbon  (FC)  from  the 1? ° t<>
IACP field experiments,  we can  transform the SE-multipliers  [Fig.
standard  errors  for  the  estimated  model  parameters,  given  the  °P , t
validation designs.  By way of illustration, Fig. 7 shows the data for w 0
a function of  soil-corrected  K for the Albuquerque fine particles.    ^nt $
observations yielded parameter estimates  (±SE) of 3.1 ± 1.4  (b0) and 0-* ^
0.011 (bK) ,  with s-3.0 when Eq. (2a) was  fit to the data.  (Units for bo ^s
s are /^g-WC/m  , for b^, /^g-WC/ng-K. )  If  we then apply this  value (s) °  ie
residual SD to  estimate  the  standard  errors of the optimally selected **
resua      o  esmae    e sanar  errors o    e opmay  see       ,
samples  [FSD] ,  we would  obtain for the intercept (N-9 , D-optimal, Fig- J)
SE  - 3.0 x  0.823  or 2.5  iig-C/m3,  and a central  point (K,Pb-150,75 W. }•
prediction SD  of 1.8 Mg/m  , for  the  3-parameter model (Eq. 3).  For tft^t;
parameter model  (Eq. 2a) ,  the  same  set of samples would yield  a WC-inte   \
with SE - 3.0 x  0.565 or 1.7 Mg-WC/m3.  Similarly, for  the  3-parameter ^ flf
SE  estimates for the K  and Pb coefficients would be given  by  the produ°
s with their SE-multipliers  (0.0064 and 0.0088,  resp.).
                                                                        e "f
       The  foregoing procedure  serves the  extremely  important  purp°s ^fl
obtaining  approximate  a  priori  estimates  for  the  SE's and  pre^* yjt
intervals  of the  optimal   sample  sets.    This  is,  in fact,  essentifli .$
judging  the adequacy  of  the experimental  design.    More  refined es
                                                                      t*
could  be  given,  taking  into  account  the   imprecision   of   the  se
variables, but  such  refinement is beyond the  scope of this discuss i-0*1'
in  any  case  would have  relatively little effect  on the conclusions.
attention  to such "errors- in-variables"  issues is  mandatory,  however«
proper parameter and uncertainty estimates following the experiments.

3.2. Optimal Design Applied to the Albuquerque  Data

       Looking again toward the previous field  experiment,  it is
to  perform  a  retrospective   sensitivity  analysis  for optimally
samples.  The 3  parameter  linear model for  mutagenicity using the Zu0*  ]\1
(Albuquerque) data will  serve as an  illustration.   The 44 samples
used  to  fit the  model,  R   - bo  +  K bK  +  Pb  bpD    where  R
revertants/m3  (Bernstein method  with  S9  activation,  extractable
were  reordered  using  the  (D-opt)  criterion   of  section  1.    The
showing the 95% confidence interval for the intercept (bo) vs number °
optimally selected samples analyzed,  is given in Fig. 8.  The benefit °
optimal sample  selection is  dramatic; we see  that practically  all  °
information is contained in about half of the samples.  Considering fhe
of mutagenicity testing, this would correspond  to a  substantial saving5'
                                   858
-------
                             4.  CONCLUSION

 ;     Selection of samples for measurement according  to  sound experimental
 ^s*gn principles can greatly  increase the information gained per unit cost.
 j,6n  rather  expensive  or  time-consuming  operations  are  involved,  such
 fining   is  vital  in   order  to  (a)   identify  the  most  cost-effective
 „ sUrement  (sampling) scheme,  and  (b)  assess  in  advance  the  adequacy  of
    plan.   For the (9)  optimally selected FSD  samples,  for example,  we can
      to test  the validity  of the  elemental  tracer model  to a level  of
      ng/m^,  since  the  SE  characterizing unaccounted-for WC is expected to
      t  3.0x0.59  or  1.8  pg/m^.  Past experience (Albuquerque)  provided both
    rt for  the K-tracer model  for WC (Fig.  7),  and the value  of s  needed
        SE estimate.
(Vj  ..Resolution of  the carbonaceous material  into  its  FC  and WC  components
i^a   C measurement)  has  two important benefits: (a) the  ability to  detect
k Cc°unted-for  carbon (model intercepts)  is higher for the univariate  (2-
^ ^meter)  models  [Eq's.  2]  as compared  to the  3-parameter  model [Eq.  3]
'lit^86 °^  K*PD  collinearity; and  (b)  the  unaccounted-for carbon  itself  is
^,0niatically  resolved  into  WC   and  FC   components.     The   fit of   the
|0 UclUerque  FC  data  to  Eq.  2b,  for  example,  implied a  small  additional
(^ sil source  (uncorrelated with Pb emissions).  (This suggests that  MLR  on
pf  individual  FC,  WC  components, may  be profitable.)   Details will  be
  Sented in a separate, more complete  evaluation of the 1984-85 ^C data.

                             ACKNOWLEDGMENT

t^   We  gratefully  acknowledge  data  from  C.   W.  Lewis  on   elemental
  entrations for Boise  samples,  together  with helpful discussions and the
          of  MLR on 14C-resolved components.   W. S.  Liggett  and  S.   B.
       provided  important  guidance  on experimental design.   Finally,  we
         thank J.  W.  Winchester  for  stimulating us during his  sabbatical
totWfiS  to  think about  optimal  sample  selection  strategies to increase the
  6tltial information content of receptor modeling data.

                               REFERENCES
I,
  G- A.  Klouda, L.  A. Currie, A.  E. Sheffield, S. A. Wise, B. A.  Benner,
     R.   K.   Stevens,  and  R.   G.  Merrill,  "The  source   apportionment   of
     carbonaceous  combustion products  by micro-radiocarbon  measurements
     for the Integrated Air Cancer Project,"  Proc.   1987 EPA/APCA  Symposium
     on Measurement of Toxic and  Related Air  Pollutants. APCA Publ, VIP-8,
     p.  573 (1987);  Unpublished  data (NBS Report of Analysis) 1988.

   c-  W.  Lewis,  T.  G.  Dzubay,   R. B.  Zweidinger, and  R. V.  Highsmith,
     "Sources  of fine particle organic  matter in Boise," APCA proceedings
     this volume (1988).

  ^- L.  Massart, A.  Dijkstra,  and L.  Kaufman, Evaluation and Optimization
     gf   Laboratory Methods  and  Analytical   Procedures.  Elsevier  Publ.,
     Amsterdam, 1978, pp 333f.

  ^- A.  David and H. T.  David,  ed.,  Statistics:  An Appraisal.  Iowa State
     Univ.  Press,  1984  [Sect.  Ill,  Experimental Design]; R.  A.  Thisted,
    ILlements of Statistical Computing. Chapman  and Hall,  New York,  1988.
     [See pp  99,  159 for definition and discussion of  Fisher  Information,
     including  the  fact  that  "under general  conditions"  the  variance-
    covariance matrix is the  inverse  of the  information matrix.]

   • A. Fuller, Measurement  Error Models.  Wiley,  New York, 1987.

                                   859
-------
                         OflY
300 J
300

100-
Ft-

«.
j * *
* **f +
"3$***

4-
*+
*

                         PB
               400

               360


               200

               100


                0
                 8  50 100 156 £66 £56 303
                              NIGHT

                               PS
                        a  SB
Figure  1.   Starting  Point:  Initial fine particle Boise K-Pb  data (ng/m ?'
FS -  fire  station  [intersection], EG - Elm Grove Park  [residential site]
                     DET
                            SE-fl CB05
               3B
               as-
               20
               15

               18
                5-I
                0
                 1  23456783 1011

                     SE-M
             0.11
             0.03
             9.0?
             0.Q6
             0.05
             O.EH
             0.03
             0.02
                       1 2 3 4 5 fi 7 3 ?10M

                           SE-M  CBPei
1  23456789 1011
                    0.11

                   0.0E5

                    a. as

                   0.035

                    0.01
                                        1234567E3 1011
Figure  2.   Incremental  Optimization.   The plots  show the values  of
N-3  objective  functions  vs  the 3rd  sample  selected.   The  "best"  ~~( &
(marked with boxes) correspond to a maximum for  Det(X'X) and minim3 *°
three SE-multipliers (SE-M's)  [FSD  samples].
                                     860
-------
                        DET C/E6)
                5000
                4000
                3000
                2B00
                1600
                                                SE-M  CB0>
                   345678 310111213
                       SE-M 
                                         SE-M (B0)
"bj
    t^-
               5000
               4000
               3000
               2000'
               1000-
                 0
                  345678 310111213
                      SE-M 
-------
JVV -
1 250 -
T
rt
I
U
II 100 _
M
_ 50 -
°-
8 —
C
- 	 	 * •
A
A ' •
4fr /I
A
A C'
* A
V
* O i.
• o
A A ° 0
jr1 A o
1 ' I 	 "' I ' 1 '
) 50 109 150 2E
                                   LEAD
Figure  5,    Distribution  of
periods  (9  samples each) .
triangles,  the  residential site  [EG].  Day  and night samples are
by open and filled symbols, respectively.
   C~ selected  samples  for  all  4  Boise
Circles  represent  the  "vehicular"  site     i
                                                                         S*
                                                                         [' i
                                                                         s^
40-
30-

20-
10-
0-
.1
x10*
*0-
30-
20-
10-




r
00



j. * *
50-




—



^



n K' 4°"
i 30"

"Un
h\jn




f"
n Pb

•
L... .
0 100 200 300 400 SOP "-SO O SO lOO ISO 200 2SO 300 3SO 400
CONCENTRATION CONCENTRATION
2JOn
* 180-
*>
130-
-
80-
30-
»»•*
TII . 	 .10-

1
I
' * '
'.-1 .4 >
1- ' ». ,
, ,. J J ••
              0   5    10  15   20   25  JO
                NUMBER OF SAMPLES IN CALIBRATION
                0  20 40 80 «0 100 120 MO ISO 180
                           Pb
                                                                          „!•»
Figure  6.    Selection of  mutagenicity  model calibration samples.   ^all"v,t)
(upper  left)  shows the  K-histogram for  all  samples.   Panel-b (upper * °yfi
shows  the  Pb-histogram.    Panel-c  shows  the Det(X'X)  vs number  of sa^ ^
selected  [FSD-only].    Panel-d  shows   the  sample  distribution   [FSD]
selection codes.
                                     862
-------
                               IACP (ALBUQUERQUE:) 1985
68-

58 -
U
0 «B-
0
D
c 30-

20-
18-


X
»»
X
X
J*f
X
x X
X'x
X**
X
X
x*x
X
1 . 1 I 1 . I 1 I ,




U
G
M





8 5Pi 188 158 288 858 389
                                  K (CORK)
  lgure 7.  Wood  carbon vs soil-corrected K for Albuquerque (IACP 1984-85)
                        CONSTANT TERM AND C,l. FOR ZUNI PARK DATA
                       too
                       so -
                        0 -
                      -so-
                     -100
                         0      10     20     JO      40     50

                         NUMBER OF SAMPLES IN CALIBRATION SET
r 1
  gure 8.    Retrospective  Dopt  design  applied  to  Zuni Park (Albuquerque)
8 Cagenicity  data.    The abscissa shows  the 95%  CI  for  the  intercept  (b )
  fctlar patterns were obtained for  the  bK and bp^  intervals.               °
                                    863
-------
SOURCES OF FINE PARTICLE ORGANIC  MATTER
IN BOISE
Charles W. Lewis, Thomas G. Dzubay,
Roy B. Zweidinger and V. Ross Highsmith
U. S. Environmental Protection Agency
Research Triangle Park, NC 27711
     Ambient concentrations of fine particle extracted organic matter
measured at  the  Elm Grove Park  and  Fire Station sites  in Boise have
apportioned  to  their  two  principal sources,  woodsmoke and  motor ve
emissions.  A multiple linear regression method using lead and potassium-
tracers for motor vehicles and woodsmoke, respectively, was employed i°  nt
source apportionment.   On average woodsmoke  was found  to be the domi^f
contributor to EOH at both sites  and  during  both day  and  night.   In spite,c|j
the 90 % reduction in the lead content  of leaded gasoline in the  U.S., vtl
has been in  effect  since January  1986,  lead  still appears to  be a
tracer of motor vehicle emissions.
                                   864
-------
 Production

 HQ,  ?-n airsheds whose atmospheric mass loadings are dominated by woodsmoke and
 t^f    source  emissions,  extractable organic matter (EOM)  generally consti-
  €? ~*® ^ or more °* tne fine particle ambient mass concentration.  The EOM
  *    the portion of tne amt>ient aerosol which is associated with mutagen-
 Na      F°r  k°th reasons it  is  of  interest to determine the  percentage  of
 s0uSUred  EOM which  is  separately  contributed  by  each  of  these  expected
 4}Ltces» as  well as  any other less obvious ones.   In past IACP  studies  in
 *tidUj?UerqUe and Ralel#n a multiple linear regression (MLR) approach, using Pb
 L  "-based tracers for motor vehicle and woodsmoke sources, respectively,  has
 Sov  Very successful in  the source apportionment of EOM  (and mutagenicity) .
 chafiver' since 1985 when  the  previous studies were conducted,  a significant
    6 haS occurred in  tne  pb emissions  of motor vehicles,   related  to  the
foil  Vide  9°  * reduction  in  the  Pb content  of leaded gasoline.  In  the
tgc     g we Present source apportionment results for EOM measured during the
tycen-t IACP Boise study, with particular attention  to the current viability of
  as a tracer of motor vehicle emissions.
pt  *ield sampling and sample analyses followed closely the methods used in
ta   Us IACP studies1'2.   Samples were collected at a residential site (Elm
     Park " EGP^ and a roadway site (Fire Station - FS)  from November 1986 to
        1987-  Samples for EOM analysis were collected on Teflon-coated glass
C(J er filters,  using multiple Hi-vol  samplers  equipped with impactors whose
s  Point was  2.5 urn aerodynamic dia.  A dichotomous sampler collected fine (0
^,•5 urn  dia)  and coarse  (2.5  - 10  urn  dia)  particles on  separate  Teflon
fay fts  wnich were subsequently analyzed for trace elemental composition by x-
|Jj fluorescence.  All  samplers were  operated  on  the same 12-h day (7 am - 7
Of J*nd  night schedule.  From the 60 -  70 sampling periods available in each
 ft "je four site - diurnal categories  (two sites,  day and night), chemometric
 ^derations3  were used  to  choose  25 each for EOM measurements. For  the
        100 sampling  periods  EOM measurements  were  performed  following
       extraction with dichloromethane.

 tePtor Modeling

    The source apportionment of  EOM was performed  with  a single element
  Cer MLR model of  the form

    EOMt = a (Woodsmoke  tracer)^ + b  (Motor vehicle  tracer)i  + c        (1)

     the subscripted  quantities  are  measured  simultaneously during each
    Lng period  i, and  the initially  unknown coefficients a, b, and c,  are
         by MLR  using data  from all  the  sampling periods together.   The
        concentration  of  a tracer  for sampling period i multiplied by  its
          is the  best estimate of the corresponding source's contribution to
^   —'" concentration during that sampling period.  The  intercept  c can be
ia^d either as  the average contribution of additional unknown sources, or
   *"~asure of inadequacy of a two-source  model.  For either interpretation c
     be  small  (relative  to  the  average value  of EOM)  for an acceptable


   AS in previous work1 a  soil-corrected fine particle  potassium tracer  was
       woodsmoke:

       = K± -  P8i  (K/Fe)soil                                          (2)

        soil ratio was 0.72  ± 0.09 (mean ±  std dev),  as determined from  the
      K/Fe  ratio in the  coarse dichotomous samples.  The  same ratio  was

                                   865
-------
found at both sites.  Previous work in Denver4, Albuquerque1 and Raleigh3 has
given  soil  ratios  of  0.68,  0.45,  and 0.42,  respectively,  with standard
deviations in the range of 10 - 15 JC.   Soil-corrected potassium calculated in
this  manner  has been  shown  to  be equivalent to  water-soluble  potassium5.

     The best indication  that Pb and  Br are motor-vehicle related  is a Pb/Br
ratio  near three for  fine particles (summertime acidic  atmospheres  often
result in larger and more variable ratios).  Figure 1 shows fine particle Pb
vs Br  for  samples  collected during the IACP study in  1985 in a residential
neighborhood  in Albuquerque.   Figure 2 shows the  corresponding  result for
Boise  based  on  samples from both  the  EGP and FS  sites.   While the scale of
Figure 2 is a factor of five smaller than Figure 1, the  slopes  are similar and
near  the  expected  value.   Figure 2  suggests  that Pb  (or  Br)  will  be an
adequate motor  vehicle tracer in Boise, although somewhat more noisy than in
earlier work.

Results

     Table I shows  average  values of  the  tracers measured at  the two sites,
separately  for day and  night,  based on  the  100  periods  for   which  EOM
measurements were  also available.  The pattern is as anticipated:  (1)  the
concentration of the woodsmoke tracer (K*) was higher at night  than during the
day, and was higher at the residential site  than  at the roadway site, and (2)
the Concentration of the motor vehicle tracer  (Pb)  was slightly higher during
the  day  than at  night,  and  was  higher  at  the roadway  site than at  the
residential site.

     Table II shows the results from  the  equation

     EOM.  =  (117 ±  8)  K*.  +  (66 ± 16) Pb.  +  1.7 ± 1.1                  (3)
     r = 0.89,  n =  97

which is based on fitting equation 1 to data from both sites  together.  Of 100
cases  originally  available,  three were omitted  in  the fit  represented by
equation 3.   All  three were from the EGP site,  the  two highest measured
nighttime concentrations and one small daytime concentration.   Including these
three cases increased both a and b by  only 10 %, but with very high residuals
for the three cases.  Neither were included in calculating any of the values
shown in Table  II.   There is seen  to be generally good agreement between the
measured EOK values and the totals resulting from the sums of  the first three
columns.  The dominant  contribution of woodsmoke at both sites is apparent.

     Using Br   rather  than  Pb gave source  apportionment results  generally
similar to  those  shown in  Table II,   with mostly  the  same cases  as  before
emerging as outliers.   Treating the two sites separately however gave less
consistent results.  Thus while the Pb  coefficients (with their uncertainties)
found separately for the two sites overlapped that found for the combined data
set,  the Br  coefficients did not.   For  both Pb  and  Br the  separate site
analyses gave source apportionment results which increased the motor vehicle
contribution at the EGP site  relative to the FS  site.   For  all  analyses,
however, the woodsmoke  contributions  were relatively unchanged and remained
dominant.

Conclusion

     In spite of the reduced ambient concentrations of Pb since January 1986,
the Boise  data  suggests this  element is still  an adequate  (although more
noisy) tracer of motor  vehicle emissions.  An MLR receptor model using Pb and
soil-corrected K showed an overall dominance of woodsmoke over motor vehicles
in their contributions to ambient EOM at both the residential and  roadway
sites.  Remaining ambiguities in the  source apportionment  results  should be

                                   866
-------
resolvable vith 14C analyses  that are planned.

    Although research described in this article has been funded by the U.S.
Environmental Protection Agency, it has not been subjected  to agency  review
and,  therefore,  does  not necessarily  reflect  views of  the Agency, and no
official endorsement should be inferred.

                                 References

1.  C.W. Lewis, R.E. Baumgardner, R.K. Stevens, L.D. Claxton, J.  Lewtas,  "The
    contribution of woodsmoke and motor vehicle emissions to ambient ae'rosol
    mutagenicity," Envir. Sci. Technol., in press  (1988).

2.  R.K. Stevens, C.W.  Lewis,  T.G. Dzubay, R.E.  Baumgardner,  R.B. Zweidinger,
    V.R. Highsmith, L.T. Cupitt, J.  Lewtas,  L.D.  Claxton,  L. Currie,  G.A.
    Klouda, B. Zak, "Mutagenic atmospheric  aerosol sources apportioned by
    receptor modeling," Proceedings, ASTM Boulder Conference, July, 1987 (in
    press).

3.   L.A. Currie, K.R.  Beebe, G.A*  Klouda, "What should we  measure? aerosol
    data: past and future," EPA/APCA Symposium on  Measurement of  Toxic and
    Related Air Pollutants, Raleigh, NC, May 1988.

4.   C.W. Lewis, W. Einfeld, "Origins of carbonaceous aerosol in Denver and
    Albuquerque during winter," Envir. Internat. 11:243  (1985).

5.   C.P. Calloway, S.  Li, J.W. Buchanan, R.K. Stevens, "A refinement of the
    potassium  tracer  method  for  residential  woodsmoke,"  Atmos. Envir.,
    submitted for publication (1988).
                                   867
-------
Table I.  Average Measured  Concentrations  (ng/m3),
                        Elm Grove  Park
                                                      Fire Station
0700 - 1900
1900 - 0700
                 K*
                 95
                 155
                  Pb
                  50
                  43
                         K*
                         70
                        138
                             pb
                             80
                             68
Table II. Average Contributions  to Boise Ambient  EOM  Concentrations (V&
0700
1900
1900
0700
0700
1900
1900
0700
    Wood
11.6 ± 1.2
16.8 ± 1.7
    Wood
 8.2 ± 0.8
16.1 ± 1.6
                                      Elm Grove  Park
  Mobile
3.4 ± 0.9
2.5 ± 0.6
  Other
1.7 ± 1.1
1.7 ± 1.1
       Fire Station

  Mobile      Other
5.3 ± 1.3   1.7 ± 1.1
4.5 ± 1.1   1.7 ± 1.1
Total
16.7
21.0
            Total
            15.2
            22.3
16.3
23-8
             13.J
             21-9
                                    868
-------
             Albuquerque  1985
                                           -rf"'
                                       X' D
                                      =  32 + 2.4*Br
 8ure 1.  Fine particle Pb vs Br in Albuquerque, 1985.
                                                400
1
i
 •30f
 280
 230
 240
 220
 200
 180
 160
 140
 120
100
 80
 60
 40
 20
 0
              Boise 1986-1987
                                Pb =  12  +  2.2*Br
                20
40         00
 Br, ng/m3
                                            80
                                                          100
    0
    <•  Pine particle Pb vs Br in Boise, 1986 - 1987.
                           869
-------
ANNULAR DENUDER RESULTS  FROM  BOISE,  ID
R.K. Stevens, T.G. Dzubay, R. Baumgardner, Roy  Zweidinger
and Ross Highsmith, U.S. Environmental Protection Agency;
D. Lovitt, W. Ellenson, Northrop Services, Inc., R.  Hardy,
Morrison-Knudsen Engineers
     Samples from annular denuder systems  (ADS) collected during  the wiij* -
of 1986-1987 in Boise, ID have been analyzed for HN02, S02, HNO  , S04" N°'
content.  The correlations between the data from the  ADS and XRF
measurements of Pb and K in the fine particles were examined to determi116
if any of the species collected by the ADS (eg HN02,  HN03) would  serve *s
receptor modeling tracers for the extractable organic matter (EOM) from
mobile and/or woodburning sources.  Results of these  correlations will »*
discussed.  This report also discusses the precision  of the ADS         ^
measurements and compares results obtained with other measuring methods
similar species (eg XRF data for S with SO " measured with the ADS).  In ^
addition, comparison of ambient concentrations of species measured with
ADS at the two monitoring sites in Boise will also be examined in
relationship to the chemical properties of fine particles collected in   &
Boise, ID and spatial homogeneity of gases and aerosols.  Some comparis0 ^
are also made with ADS data collected in Denver and Research Triangle
NC.
                                   870
-------
                                 INTRODUCTION

 *he Annual Denuder Method (ADM)1 was used in the Integrated Air Cancer
 *roject field sampling program in Boise, ID to obtain 12 hour air quality
     on the concentrations of HNO,, S02 , HNO? , sulfate and nitrate at the
    main monitoring sites.  The ADM data will be used to determine the
         spatial homogeneity of the atmosphere between monitoring sites and
 °btain data to correlate concentrations of condensable nitroaromatic
 impounds with HN02 and HN03 .  The assumption is that these inorganic gas
 Phase nitrogenous species may be the nitrating agents that produce
 litroaromatic particles in the atmosphere.

     We will discuss in this report recent developments in sampling and
 Dialysis of atmospheric aerosols and gases using the ADM.  We also discuss
 "°v modifications in the annular denuder system may be used to obtain
 pliable measurements for HN02 and HN03 and how HN02 may cause
 cnemi luminescent N02 measurements to be overestimated.

                                 EXPERIMENTAL

     Descriptions
     Boise,  ID.  During winter of 1986-1987 air quality data and samples
     collected at two primary monitoring sites as part of the Environmental
^otection Agency's (EPA) Integrated Air Cancer Project.  One site was
*°cated in a city park (Elm Grove Park) in a residential neighborhood
^Pacted by wood smoke emissions from numerous single family dwellings.
rhe second primary site (Fire Station) was on top of a fire station
a
-------
     The objective of the experiments performed  in RTP was  to  decouple
collection of HNO  from HN02 by coating  the  first annular denuder with
0.2% solution of NaF.  Three rather  than  two denuders were  used, where
denuders 2 and 3 were coated with Na C03.  In  this configuration all  HL-
is collected on the NaF denuder and  HN02  is only collected  on  the Na2C03
denuders.  Results and implications  of this denuder arrangement are
discussed later in this report.

Summary of Sampling and Analysis Methods

     Table I lists samplers and instruments used to obtain  data on  the   .g
concentrations of major inorganic gases and aerosols in  Boise  and RTP.   j
annular denuders were extracted in 15 ml  of distilled water and stored i^
refrigerator in sealed plastic bottles.   Extracts and filters  were  shipP6
to EPA, Research Triangle Park, NC and stored  in a refrigerator prior to
anlaysis.  Filters from the ADM were sealed in plastic petri dishes and
stored in a refrigerator with the denuder extracts.  The filters were
extracted in 10 ml Ion Chromatographic (1C) Na.CO  buffer solution  in an
ultrasonic bath just prior to 1C analysis for N02 , N03~ and S04" conten

                                     RESULTS

Annular Denuder and Delated Gas-Aerosol Data

     Figure 2 is a summary of the means of daytime(AM) and  nighttime(P^/ t
data for HN02, SO , HN03f S04=» N03~» NOX and mass for samples collected
the Elm Grove Park and Fire Station  primary sites in Boise.  The sampl05 ,
were collected between November 8, 1986 and February 2,  1987 and repres6
30% of the samples collected in Boise.  The complete set of results £r0injst
the ADM will be reported elsewhere.  These samples were  collected to  aSSfljc
in receptor modeling studies to determine the origin of  extractable °r^fs
matter (EOM) and the origin of mutagenic  properties of the  EOM.  For  thi
wintertime study nitric acid was typically less  than 0.5 ug/m3 while     s
particle nitrate averaged 5 ug/m3.   This  contrasts with  summertime  stud1
in the Eastern U.S. where nitric acid is  typically 2 to  3 times the finfi
particle nitrate concentrations.1

     Table II is a summary of the correlation statistics for data colleCte<)
with the ADM systems, CO and N0x monitors and dichotomous samplers  °PejrLjr
at the two main monitoring sites in  Boise.  These correlations between *
chemical air pollutant species serve as one means of evaluating and
comparing air quality between the Fire Station and Elm Grove Sites  in
Boise.  Also these correlations are  used  in identifying  chemical specie5
which may serve as candidates for use as  tracers for receptor  modeling   j
calculations.  For example, the moderately high  correlation between Pb a
CO suggests CO may be a useful surrogate  for Pb  as a tracer for mobile
source emission in receptor modeling calculations.

     Table III shows selected ADM results from the Elm Grove Site in  Bo1
These data are presented to provide  information  on the relative daytii"6'^
nighttime concentrations of gases and aerosols and show  the variability
the amount of nitrate that evaporates from the ADM Teflon filters.

Precision and Accuracy of Sampling and Analysis  Procedures

     Two factors which are considered in  estimating the  precision of    ^
measurements obtained with the denuder system are flow rate variations
precision of 1C analysis.  In the ADM study reported by  Vossler et al
paired samples had average differences for the measurement  of  SO , HNOj*
and HN02 of 3.4%,  6.4% and 8.3%.  These rather small deviations were  duej(jn
mainly to variations in sampling flow rate of the paired assemblies

                                   872
-------
  hromatography analysis.   Typically variations were less than ± 5% for flow
  «a less than ± 3% for 1C analysis.  For our studies in Boise and RTF, NC
  «e ic analysis precision for SO,', NO ', NO ', NO ~ was typically better
  'ian + 2%,  and flow rates were maintained to ± 5%,

 §Hl|ur-Sulfate Relationship

      In Boise fine particle samples were collected with dichotomous
  ampiers to obtain mass and elemental composition data by XRF procedures
  or use in  receptor model calculations.   The mass ratio of fine particle
  «ifate concentrations measured with the ADM to the sulfur measured from
  «e dichotomous sampler by XRF for samples collected at both main
  onitoring  sides in Boise was SO " (ADM)XS(XRF) = 2.99 ± 0.24 at the 95%
  oniidence  limits.   This  is in close agreement to the SO "/S data obtained
 J  Lewis et  al  in a wintertime,  1982,  receptor modeling study in Denver.
 pje ratio of 3.00 corresponds to all sulfur being ?.n the form of sulfate.
  igure  3 is  a weighted least  squares plot of SO " versus the S in fine
  Articles sampled in Boise.   In the weighted least  squares  calculation3  the
 Va  fession line was  forced  through zero  and data weighted so that the
 fiance of  Y (SO "  concentration)  was proportional  to X2  (X = S
  °lcentration).   Thus  the  best  estimate  of  the regression coefficient is
 n"e average  of  n  slopes obtained  from each  pair of observations  Y./X..  A
 StHf   °f data  P°ints  are  be]-ow the best  straight  line fit  of the data for
 fo     concentrations  >!'2  ug m~3.   This  suggests  the  data  may exist  in two
 *e ?S*  Additional samples are being analyzed to determine whether  this is
 to   °r due to other factors such as incomplete extraction of the S04" due
 f^levated concentrations of extractable organic matter present on the


                                  DISCUSSION

     For studies performed in Boise and a previous Denver receptor study
    ormed in 1982 ,  a number of similarities in air quality data were
    rved.  For example, the mean concentration of nitrate in Boise was
    cally 13% of the fine particle mass (Figure 2).  In the previous
   :ertime study in  Denver2 nitrate represented 14% of the daytime fine
    *-"•   mass.
     *n B°*se»  HN02  concentrations averaged nearly 10% of the NO  values.
     during the day  HN02  averaged 1% of the NO,  concentration.  During
        HNOZ  daytime concentrations are typically less than 1 ug/m3.
       ,  during winter,  cold and cloudy conditions tend to reduce the
c °t°-disassociation of  HNO, ,  which may account  for a mean daytime
*dd?entration of HNO?  of  3 u«/m3'  observed in Boise (Figure 3).   In
&    ,on'  ^l  should  be noted that  chemiluminescent N0  monitors  respond to
     ,                                                x
   2  just as  it  responds  to  NO .   Therefore,  N02  measurements in Boise or
   otner ur^an  area  would tend  to be  overestimated  due to  the positive
^ e^ference  from  HNO .   This,  in  turn,  would require adjustment in
^"^luminescent N02  data.   For  example  on  January  14,  1987  the average
ton ime HN°2  concentration was 4-3 ug/m3, which was  9.5% of  the N02
^ Gentration.  Therefore the  average  daytime N02 concentration was really
in- ^8/m  rather  than  the  45  ug/m3  measured  by the chemiluminescent  NO
  nUor.                                                             *

H     Figure 2 shows data  that  indicate that SO , nitrate, fine article
     a   sulfate mean concentrations were nearly identical for data
     ted at  the two  PrimarY monitoring  sites  in Boise.   This is an
         observatlonf if calculations on pollutant  exposure  are to be
         using the data  from  the  two primary  monitoring sites.

    To improve the reliability of HN03  and HN02 data obtained with the

                                  87.?
-------
annular denuder, a modification of the annular denuder coating procedure
was tested.  In previous annular denuder studies some oxidation of HN02
nitrate on the denuder surface during sampling was possible during
summertime conditions.  Recently Febo et al4 reported on the use of a
coated denuder followed by Na,C03 denuders to decouple the collection
HNO  from HN02.   In this configuration only HNO  (and some S02) is
collected on the NaCl coated denuder, and HNO  is collected on the Na-COj
denuder.  Thus if HN02 is oxidized on the Na2CO  denuder there would 06 "
interference with HN03 measurements since the HN03 had been removed by *
NaCl coated denuder.  In our laboratory we choose to use a NaF coated    -
denuder instead of the NaCl because the Cl~ peak tended to obscure the " I
and N03" peaks during ion chromatographic analysis of the denuder extra0
In our studies the first denuder is coated with a 50:50 water:methanol   j
solution containing 0.2% NaF and 0.2% glycerine using a procedure descf*
by Vossler et al.1.  The NaF coated denuder collects 99% of the HNO, an<1 ^
some S02 at a flow rate of 10 L/minute.  The amount of S02 collected on
NaF denuder appears to depend on relative humidity.  There was no evide^
of HN02 collection on the NaF coated denuder.  The HN02 and balance of
ambient concentrations of S02 are found on the first Na2C03 denuder.
These preliminary studies suggest the annular denuder system using a
coated denuder followed by two Na2C03 coated denuders eliminates the
potential problem of HNO, interfering with HN03 measurements due to
oxidation of N02~ to N03  on the denuder.

                                    SUMMARY

     Annular denuder and related air quality data collected in Boise, ^ ^
part of the EPA Integrated Air Cancer indicated fine particle nitrate t°,J
a major fraction (13%) of fine particle mass, while sulfate represented
of fine particle mass.  Nitrous acid concentrations averaged 10% of N02
concentration and would produce a significant interference with
chemiluminescent N02 determinations.  Annular denuder data showed     .^$
substantial evaporation of fine particle nitrate occurs from Teflon fi*• *$
even during wintertime conditions.  This suggests filter pack measuremej 9
of nitrate and HN03 may have substantial uncertainties.  Introduction °
NaF annular denuder between the inlet and a Na2C03 denuder eliminates t
potential interference of HN02 with HN03 measurements and improves the  flf
annular denuder method as a tool to measure gas and aerosol phase oxide
nitrogen.
                               ACKNOWLEDGEMENTS

       The authors wish to acknowledge Stella Mousmoules for typing the ^
manuscript, J. Wu of NSI for assisting in the statistical analysis, and
Lewis for his critique of this report.

                                  REFERENCES

1.  T.L. Vossler, R.K. Stevens, R.J. Paur, R.E. Baumgardner, J.P.
    "Evaluation of improved inlets and annular denuder systems to
    inorganic air pollutants," Atmospheric Environment (In Press).

2.  C.W. Lewis,  R.E. Baumgardner, R.K. Stevens, "Receptor modeling stud?
    of Denver winter haze," Environ. Sci. Technol., 20; 1126 (1986).

3.  N.R. Draper and H. Smith, "Applied Regression Analysis," 2nd Editi°m
    John Wiley and Sons Inc., New York, N.Y. (1981).

4.  A. Febo, F.  De Santis, C. Perrino, "Measurement of nitrous and nitr
    acid by means of annular denuders," Proceed, of the IV European
    Symposium STRESA (1986).

                                   874
-------
 Table I.  Samplers and Analysis Procedures  Used  in  Boise Study

     Sampler/Analyzer
     Flow Rate
     L min"1      Species*
                               ReT
 Annular Denuder Method (ADM)   15
 (Boise, ID)


 Annular Denuder Method (ADM)   10
 (RTF,  NC)
 Anderson Inc.
 Oichotomous Sampler

 janitor  Laboratories
 ^nemiluminescent
 N°x Analyzer

 Jeckman  Gas Cell
 °0 Monitor
         16


          1
             S02, NO ,
             SO -, HN02
             HN03

             SO , HN03,
             HNO,, N03-,
             SO/

             Mass, Ele-
             mental
             NO,  NO,,  N0x
                   ft     A
                     CO
                         2


                         2
  47 mm Gelman Teflon  filters  followed by 47 mm Nylasorb filters were used
  *th the ADM system;  ADM  system fabricated by University Research Glassware,
  atrboro, NC.  A pair of  37 mm  Gelman Teflon filters was used to collect
 Particles with the dichotomous  sampler.


  able II.  Summary of  Correlation  Statistics for  Gas and Aerosol Samples
 Collected in Boise, ID

 SPecies (Site)
N
AM
N
PM
N
Overall
52 vs N02 (EGP)
52 vs NO (EGP)
)2 vs N02 (FS)
>2 vs NO (FS)
vs CO (EGP)
VS CO (FS)
	
19
19
17
17
20
18
0.700
0.570
0.381
0.610
0.718
0.920
25
25
22
22
23
24
0.564
0.634
0.634
0.858
0.643
0.744
44
44
39
39
43
42
0.552
0.529
0.375
0.557
0.682
0.835
>b
AM
 n tefers  to  data collected between 7 AM and 7 PM, PM refers to data
 Elected  between 7  PM and 7 AM.
                                  875
-------
Table III.  Annular Denuder Method Data  for Boise
DATE
SAMPLED
TIME
Concentration, ug m~3
GASES PARTICLES
HN02 HN03 S02 N03~ S04" % Nitrate
Found on
Nylon Filtef
^-*
12-6-86
12-7-86
12-7-86
12-9-86
12-10-86
12-11-86
12-11-86
12-12-86
night
day
night
night
night
day
night
day
6.95
2.00
5.35
3.29
3.83
1.37
7.52
2.79
0.0
0.25
0.03
0.16
0.12
0.34
0.09
0.20
5.40
2.55
6.66
10.53
9.66
7.60
12.03
10.62
1.68
2.40
1.46
3,05
4.32
4.70
4.46
4.14
1.16
2.04
1.88
1.39
2.12
2.08
2.91
2.41
30
81
25
56
26
40
14
29
__, 	 --
Samples were collected between 7AM-7M (day period) and 7PM-AM (night Pfirl
                                  876
-------
             ALUMINUM
                                     2-STAGE
                                      FILTER
                                  TEFLON
                                    RING
               SEAL CAP
                 COUPLER
DIFFERENTIAL
FLOW
CONTROLLER
r
^
                                                 PUMP
                                        FLOWSTRAIGHTENER
                                            SPACE
             SEAL CAP
            TEFLON RING
4mm ORIFICE
                                   1 mm
                                  ANNULAR
                                   SPACE
                                     GREASED
                                  IMPACTOR STAGE
                   SEAL CAP
                           COUPLER
             TEFLON COATED
             IMPACTOR INLET
        1
TEFLON COATED
GLASS IMPACTOR
    INLET
 BOISE, ID ADM SAMPLER
DENVER, CO ADM SAMPLER
  1 - Diagram  of annular denuder  systems used  in  the Boise, ID
EPA's Integrated Air Cancer  Project and Denver  Air Toxic Study,
                             877
-------
   16
   14
   12
 Z i
 Q


 <  8
 oc •

 Z !
 Ill
 u ! 6

 o '
 u
          ELM GROVE PARK



MASS CONCENTRATION - MASS x 10

NOX CONCENTRATION » NOX x 10
                                                            FIRE STATION
 CM CM  CO fO  T
O O O  O O
2 c/> z  z | u>

IJ ,  XI

•	AM —
                          CM  . Z : Z "
                                 !5
                                 <
es  M o  e») »t  x *"
O  O O O O 0  tJ
2  M Z z w 2  S
xi      *  5
                                -PM-
                                                    •AM-
 CM CM n  
-------
 utagenicity of Organlcs Associated with PM2.5 and PM10 HiVol
         from a Wood Smoke Impacted Residential Area
][• Watts
palth Effects Research Laboratory
:•• Cupitt and R. Zweidinger
^mospheric Sciences Research Laboratory
^search Triangle Park, NC 27711
    Particulate filter samples were collected from collocated PM10
        with and without internal impactors, thereby producing parallel 0-
 •5 and 0-10 micron samples.  The filters were comparatively analyzed for
 ^tractable organics, mutagenlcity concentration, and mutagenic potency of
 ^tracted organics.  These ambient air samples were collected on Pallflex
 Alters during the 1985 IACP wood smoke study conducted in a residential
 tea of Raleigh, NC.  Eleven sample sets representing 11 sampling periods
 6te selected for each type of sampler.  These sampling periods had fine
 rticle ,(0-2.5 micron) mass concentrations ranging from 11-129 micrograms
   cubic meter.  Duplicate filters from like samplers which were collected
      the lowest loading periods were pooled to gain sufficient sample for
 oassay testing.  This particle size comparison study indicates that
 ^tractable organics and organic mutagens were primarily associated with
 fte 0-2.5 micron particles.

 ^reduction

*   An ambient air sampling program was conducted by EPA's "Integrated Air
 *ncer Project" (IACP)  in a wood smoke Impacted residential neighborhood
j* Raleigh, NC during the winter of 1985.  One of the goals of that initial
 *VCp effort was to characterize organics adsorbed on particles collected by
 fte various samplers used.  This particular study of those samples was
 dertaken to examine the amount of organics and the mutagenicity of
 *fctactable organics associated with the 0-10 and 0-2.5 micron diameter,
 'tides respectively collected by PM10 and PM2.5 samplers.  The Ames
          typhimuriura histidine reversion assay was used to evaluate
                                  879
-------
Experimental

     Study sets were assembled with each set containing filters from
and PM2.5 size selective inlet collocated HiVol samplers.  Filters with*11
set were all from the same 12 h sampling period.  Eleven such sample sets
were compiled from 11 sampling periods with fine particle (0-2.5 micron)  .
mass concentrations ranging from 11 to 129 ug/m  and fine plus coarse ("'
micron) particle concentrations ranging from 16 to 136 ug/m  ,  The PM2.5
filters (1-6) within a set were pooled and extracted together.  PMlO    ,
                                                                        '
filters (1-5) of the same set were similarly treated.  The number of ---  -
and PMlO filters and, consequently, the amount of extractable organics
each pool within a set were matched as closely as possible.  All 11 sets
consisting of the 22 pooled samples were analyzed together for
determination of extractable organic mass (EOM) and mutagenicity.

                    Extraction and Determination of EOM

     Filters from the same set and sampler type were pooled and
extracted for 24 h with dichloromethane.  Extracts were filtered (0.2
micron filter),  concentrated by rotary evaporation, and volume adjusted
exactly 10 ml.  Duplicate determinations of EOM were made on 0.25 ml
aliquots from each pooled sample.  Aliquots of the stock solutions we£e
then solvent exchanged to 5 ml dimethyl sulfoxide (DMSO) to give
extractable organics concentrations of approx 2.5 rag/ml for bioassay.

                    Bioassay Analysis for Mutagenicity

     Samples in DMSO were assayed for mutagenicity in the Salmonella
typhimurium histidine reversion assay with strain TA 98 .  The specif^c *
test protocol followed the procedures originally described by Ames et a
and revised by Maron and Ames .   Sample sets were tested simultaneously .$
within the same experiment to minimize bioassay variation.  Each sampl6 $
diluted with DMSO to 2 mg/ml and tested at six doses with duplicate P^&o0^S
and with aroclor induced S9 metabolic activation at each dose.  Spoilt
reversion rates for TA 98 were 25-50 colonies per plate after a 72 h
incubation.  A set of positive controls were incorporated in each
experiment.  The data were analyzed by the statistical methods of
et al. .   Slope values were reported as revertants/ug (rev/ug) of
extractable organics.

Results and Discussion

     Apportionment of the organic contribution of the sample sets,
according to the methods of Stevens et al. ,  indicated that wood smoke
contributed between 88 and 97% of all the organic compounds present.
Characterization of the samples collected by the PMlO and PM2.5 HiVo^
samplers was accomplished by respective determinations of organics/m »
rev/m , and rev/ug of extractable organics.  Statistically significant
differences were found for each of these parameters with the PMlO val°e
being a few percent larger.  The differences in rev/m  and ug of
organics/m  correlated positively with the total mass loading of each   ^
sample set.  The differences in rev/ug of extracted organics diminish6  es
the percent fine fraction in the PMlO sample increased.  The latter va
were calculated from PMlO dicot sampler fine and coarse particle
measurements for each sampling period.  It should be noted that as the
                                   880
-------
        fine mass  in  ambient air  increases  (x axis  for Figures  1  and  2),
   particles collected by  the  two samplers theoretically become  more and
     alike until they are identical at  100%.
                                                3
     Figure 1 shows that the amount of  organics/m   that were collected by
   two  samplers was  nearly identical for several periods.  The PM10
s&mpler, however,  collected more  organics for 8 of  the 11 periods with an
&Verage  percentage difference of  7.7 +/- 13.8.  The differences increased
6s  the total mass  loading increased.
             3
     The rev/m  rautagenicity comparisons in the Figure 2 plots  show
telationships that closely  parallel those of Figure 1.  The PM10  values
 re  higher than the  PM2.5  values for all 11 periods with an average
Percent  difference of 20.0  +/-  13.3.  Figures 1 and 2 demonstrate that
*xtractable organics  and mutagenicity were primarily associated with  the
     particles.  This finding is  in agreement with  the conclusion of
      h who reported that ambient air mutagens appear to be present in
      concentrations in the smallest particles (<  2 microns) than in
      particles.
    Figure 3 shows potency values for organics collected by  the^two  types
°*  samplers derived from plots of rev/m  versus ug of organics/m   for each
Sai"pler.  The slopes of the linear regression lines yielded a PM10 potency
of  0.56 +/- °-03 rev/ug and a PM2.5 value of 0.55 +/- 0.03 rev/ug.  The  r
 alue for each plot was 0.98 and 0.99 respectively.  These potency
 ^terminations are very comparable to a similarly derived potency  value  of
 •61 +/- 0.05 for PM2.5 organics collected during ambient air sampling of a
 °°d smoke impacted neighborhood of Juneau, Alaska .

    An estimation of  the potency of organics associated with only the
      fraction (2.5-10 micron) particles was obtained by plotting the PM10
     PM2.5 rev/m  differences against the same ug/m  differences  (plots
   shown).  The linear regression slope showed a potency of  0.55  -f/-  0.06
 ev/ug.  This indirect manner of arriving at a coarse particle potency
 *lue resulted in considerable scatter of data points; however, an r  value
°f  0.95 was obtained.

Delusions

    For this limited  set of wood smoke dominated samples, extractable
.'••Sanies and mutagenicity were found to be primarily associated with  the 0-
 l5 micron particles.  The PM10 sampler, however, collected more organics
   8 of 11 periods and more mutagenicity for all 11 periods.  The observed
       are most easily explained by assuming that the ambient distribution
   Particles with extractable organic compounds extends beyond the cut
   t of the PM2.5 sampler and that this small additional fraction (approx.
   is collected by the PM10 sampler.

    The potency of the organics collected from this Raleigh, NC wood smoke
        neighborhood was 0.56 rev/ug for the PM10 sampler and 0.55 rev/ug
j.* the PM2.5 sampler.  These values are nearly identical to  the 0.61
 6v/ug potency found in PM2.5 organics collected from a wood  smoke impacted
^ghborhood of Juneau, AK.  More definitive experiments, however, should
 e  conducted to better characterize the mutagenic potencies of the fine  and
      fractions.
                                  881
-------
Disclaimer

     The research described in this article has been reviewed by the He*1
Effects Research Laboratory, U. S. Environmental Protection Agency and
approved for publication.  Approval does not signify that the contents
necessarily reflect the views and policies of the Agency nor does mend011
of trade names or commercial products constitute endorsement or
recommendation for use.

References

1.  J, Lewtas and L. Cupitt, "The integrated air cancer project: progr*"1 ,
    overview",  1987 EPA/APCA Symposium on Measurement of Toxic and RelS^"
    Air Pollutants, Research Triangle Park, NC.

2.  L. D. Claxton, K. L. Dearfield, R. J. Spanggord, E. S. Riccio, and K'
    Mortelmans, "Comparative mutagenicity of halogenated pyridines in *-"
    salmonella typhimurium/mammalian microsome test", Mutat. Res.( in
    press.

3.  B. N. Ames, J. McCann, and E.  Yamaski, "Methods for detecting
    carcinogens and mutagens with the salmonella/mammalian-microsonie
    mutagenicity test", Mutat. Res.,  31, 347 (1975).

4.  D. M. Maron and B. N. Ames, "Revised methods for the salmonella
    mutagenicity test", Mutat. Res.,  113, 173 (1983).

5.  L. Berstein,  J. Kaldor, J. McCann, M. C. Pike, "An empirical appr°aC
    to the statistical analysis of mutagenesis data from the Salmonell*
    test".  Mutat.  Res., 97, 267 (1982).

6.  R. K. Stevens, C. W. Lewis, T. G. Dzubay, R. E. Baumgardner, L. T.
    Cupitt, V.  R.  Highsmith, J. Lewtas,  L. D. Claxton,  B. Zak, and L. ^a
    Currie, "Source apportionment of mutagenic activity of fine
    collected in Raleigh, NC and Albuquerque, NM", 1987 EPA/APCA
    on Measurement of Toxic and Related Air Pollutants, Research
    Park, NC (May 1987).

7.  J. L. Huisingh, "Bioassay of particulate organic matter from ambi^11
    air", Short Term Bioassays in the Analysis of Complex Environmental
    Mixtures II,  Plenum Publishing Corp., 1981, pp. 9-19.

8.  R. R. Watts,  R. J. Drago,  R.  G. Merrill, R. W. Williams,  E.  Perry,
    J. Lewtas,  "Wood smoke impacted air: mutagenicity and chemical
    of ambient  air from residential neighborhoods" , Proceedings: 1987.   ^
    Annual Meeting, Pacific Northwest International Section,  Air "-11llti-x
    Control Association, Seattle,  WA.
                                   882
-------
 0)
.a
u
o

OT
.o
'E
o
            1	1	1—T	1	r—i	
                % fine mass in PM10 dicot

                    •  PM10     +   PM2.5


Figure  1.    PM10 and PM2.5 organics/cubic meter
 0)
•s
 E
 o
70




60



50



40




30-




20-




10-
      62
Figure 2.
              fs
               ITT
                                             94
66    70    74   78    82    86    90

     % fine mass in PM10  dicot
        •  PM10    +  PM2.5

PM10 and PM2.5 revertants/cubic meter
                         883
-------
REVERTANTS/CUBIC METER
        70

        60

        50

        40

        30

        20

        10
SLOPE - REV./US POTENCY
PM10
                                                     REG!**5
                                                PM2.5
          0   10

 PM10 SLOPE-0.56
 PM2.5 SLOPE-0.55
     20   30   40   50   60   70  60   90   100  110
          UG ORGANICS/CUBIC METER
 Figure  3.   Potency  of  organics
                                  884
-------
^ransformation of Boise Sources:  The Production and Distribution of
"utagenic Compounds in Wood Smoke and Auto Exhaust
>• T. Cupitt
ionospheric Sciences Research Laboratory
h* D. Claxton
tl^alth Effects Research Laboratory
 5 EPA
>esearch Triangle Park, NC 27711 and
Hi' E- Kleindienst, D. F. Smith, and P. B. Shepson
R°rthrop Services, Inc.
 Search Triangle Park, NC 27709

         Emissions from the principal combustion sources in Boise (e.g.,
     vehicles and wood smoke) have been shown to produce a wide variety
  routagenic compounds.  These mutagenic species are associated with both
p ' Particulate-bound and the vapor phases.  They are formed both as
iu!n!ary pollutants, emitted directly from the source, and as secondary
        produced from normal atmospheric photochemical processes.  The
     ve potency df the mutagenic compounds in the gas and aerosol phases
  n9es as atmospheric transformations occur.  In the source emissions,
    of the mutagenicity is associated with the particle-bound organics.
  adiation of the dilute exhaust materials in photochemical simulation
         however. Increases the mutagenicity of the gas-phase pollutants
^°stantially, suggesting that the overwhelming majority of the mutagenic
 Ma" may exist in the vapor phase.  The changes in chemistry and
            which occur during irradiation have been characterized in an
       to identify the mutagenic compounds present.
c
 na
                                  885
-------
     A series of atmospheric simulation experiments have been conducted
to examine the effects of photochemical processes on the lifetime and
fate of wood smoke and auto exhaust, as part of the Integrated Air Cancer
Project (IACP).  Dilute mixtures of these complex emissions were injected
into a Teflon smog chamber and irradiated to simulate the reactions
expected in urban air.  In the transformed mixtures, the bulk of the
mutagenicity was found to be associated with the gas-phase products
rather than the aerosol-bound chemicals.  The mutagenic gas-phase
products have been shown to be stable in the smog chamber for more than
10 hours.

     One goal of the IACP is to define and improve the estimate of the
risk associated with exposure to urban air pollutants. The strategy of
the IACP has been to begin by examining the risks from exposure to
"products of incomplete combustion" (PICs).  Several attempts to estimat
the risk from exposure to urban pollution have identified "products of
incomplete combustion" (PICs) as significant contributors to the poten-
tial risk.  A variety of compounds often found in aerosol samples of P^
have been identified as mutagens and potential carcinogens.  Because of
such results, the IACP research has focused on two common combustion
emissions often found in urban areas, namely, wood smoke and motor
vehicle exhaust.  In addition to the possible presence of mutagens in tn
emissions themselves, evidence has appeared recently in the literature
indicating that reactions of species such as 03 and ^05 on the surface
of atmospheric particulate matter can lead to increases in the mutagenic
activity of the adsorbed species.

     We present the results of experiments which we carried out to
measure the chemical changes and the mutagenic activity of both the gas*
phase species and the aerosol-bound compounds.  The product mixtures
under test were produced in a Teflon smog chamber, and the Ames test
plates used for measuring the mutagenicity of the gas-phase species
dosed by continuously flowing the reaction chamber air over the uncover"6
plates, thereby permitting the soluble species to deposit continuously a
long as the plates remained uncovered.

     The irradiations were carried out in two different smog chambers.
One chamber is a 22.7 cubic meter Teflon cylinder housed in a truck
trailer, and surrounded by black lamps and sun lamps to provide the    r
necessary illumination.  The second chamber is an 8.5 cubic meter outdo0
chamber, which can be shielded from sunlight by an opaque cover when
necessary.  The emissions from an oak-burning wood stove and from the
autos under test were first injected into a dilution tunnel  to lower the
temperature of the exhaust gases and to bring the chemicals into a phase
distribution like that in ambient air.  A portion of the diluted combus'
tion emissions were then injected into the reaction chambers.  Nitroge^
oxides were sometimes added to bring the hydrocarbon/nitrogen oxide rat1
more in line with that found in urban areas.  In the wood smoke and
automobile exhaust experiments, the initial total hydrocarbon concentra"
tions were about 20 and 12 ppm Carbon, respectively.  Once the appro-
priate starting conditions were achieved, the lights were turned on (or.
the chamber was uncovered) and the reaction was allowed to proceed unti
ozone maximum was achieved.  The lights were then turned off (or the
cover was replaced) and the bacteria exposures begun.
                                   886
-------
     Four  190-L Teflon-coated exposure chambers were used for exposure of
    bacteria to the various air streams involved.  For example, in the
     smoke experiments, the bacteria were exposed (1) to the diluted wood
Snioke  (the initial reactants), (2) to the transformed wood smoke (the
Products), (3) to the clean air used to dilute the reaction chamber, and
^)  to the ambient air used in the dilution tunnel.  The testing for
^tagenicity was conducted by exposing the Ames test bacteria Salmonella
typhimurium to the test gases.  Strains TA 100 and TA 98 were used, both
 |th  and without metabolic activation.  The exposures were conducted by
Blowing the air to flow through the exposure chambers, each loaded with
Approximately 50 covered glass petri dishes containing the bacteria in a
^trient agar.  Dose response curves were generated by varying the
-xposure time for different sets of plates, and plotting the observed
      i cities against the exposure time.  Because the agar in the test
    es is mostly water, those species that are water soluble (i.e.,
      are expected to deposit in the test plates.  After exposure, the
      were incubated for 48 h and the number of revertant colonies
c°unted.

    The bacteria in the flow-through exposure chambers were assumed to
.eact only to the gas-phase mutagens, since imposition of a filter in the
 pansfer line did not change the response.  A comparison of the chemical
 Or|cent rations entering and leaving the exposure chambers showed that the
 *Posure chamber was not removing all of the gas-phase mutagens.  To
        the total mutagenicity in the  gas stream,  numerous experiments
    conducted with two exposure chambers in series.  In calculating the
     mutagenic burden of the gas phase, we assumed that the second
    er was as efficient as the first in removing vapor-phase mutagens.
     that assumption is probably not valid, our calculated values  for
   vapor-phase mutagenicity actually represent a lower limit.

    The mutagenicity of the parti cu late-bound organics was easier to
    ate.  Filter samples of known volumes were collected,  and  the
i ^genicity of the filter extracts was measured using the standard plate
 Corporation test.

U    Chemical  concentrations throughout the irradiations were monitored
- 'n9 a wide variety of techniques, including continuous gas monitors,
* s» liquid and ion chromatography, and gas chromatography/mass spec-
 r°nietry.
        cnemical  composition and mutagenic activities  of the gas-phase
HI  Particulate-phase species were measured before irradiation of the
 Xtures and after the ozone maximum was reached.   For  both  wood smoke
1^° automobile exhaust, the mutagenic activities of the gas-phase species
r^r6ased dramatically as a result of the irradiation.   To compare the
native mutagenicitles of the two phases, we have expressed the observed
p^a9enicities in  common units of revertants per cubic  meter.   The gas-
tjas3 mutagenicitles have been corrected as described above, based upon
   results of using the two exposure chambers in  series.   The dramatic
       for both wood smoke and auto exhaust are shown  in Figures 1 and
      developing  the dose response curves for the gas-phase mutagenic
         we exposed the bacteria for periods of 10 hours and  longer.
                                  887
-------
Since the dose response profiles continued to be linear throughout that
time period, we conclude that the mutagenic products must be stable in
air for periods much longer than ten hours.  These products  are also
likely to be stable for long periods of time in the ambient  atmosphere.

     Additional experiments have been conducted to help us understand the
processes involved in the production of gas-phase mutagens.   Wood smoke
was reacted with ^05 to assess the potential for nighttime  reactions.
In this experiment, we also observed a dramatic increase in  both the gas-
phase and particulate-phase mutagenicity.

     In one wood smoke experiment, an XAD-2 cartridge identical  to those
used in the lACP's Boise field study was inserted into the sample line
between the reaction chamber and the bioassay exposure chamber.   In this
case, the mutagenicity measured in the chamber was reduced considerably*
and only about 15% of the gas-phase mutagenicity was passed  through the
XAD-2 cartridge.  When the XAD-2 cartridge was extracted and bioassayed,
about 10% of the original gas-phase mutagenicity was recovered.   Clearly*
much of the mutagenic potency was lost in  this process.  The mutagenic
species may have been destroyed on the XAD-2, or lost in the extraction
and work-up of the XAD-2 sample.  Indeed,  the mutagenic species  recovered
from the XAD-2 cartridge may have been a completely different set of
chemicals altogether, with none of the gas-phase mutagens.

     Clearly, emissions, like wood smoke and auto exhaust, from common
combustion sources can contain potentially hazardous chemicals as primary
pollutants.  The organic compounds and nitrogen oxides present in those
emissions can also react photochemically to yield products which may be
much more mutagenic than the original starting materials. The potential
for the formation of gas-phase mutagens should not be overlooked.
                                   888
-------
      Comparison of Gas and Paniculate Phase
    Mutagenicity of Dilute Wood Smoke in Air
u
—
o

o
S
3
O
V
a
17500
15000
12500
10000
7500
5000
2500
0
-
—
—
j- TA100
~
—
~ __«= 	







| Gas Phase
Paniculate Phase

•
TA98
~~~\
LM___.
  BEFORE      AFTER
IRRADIATION  IRRADIATION
                                      BEFORE     AFTER
                                    IRRADIATION  IRRADIATION
  Figure 1. The effects of irradiation on the yas phase and paniculate phase mutagcnicity of diluted wood smoke,
       using two different strains of bacteria.
       Comparison of Gas and Particulate Phase
     Mutagenicity of Dilute Auto Exhaust in Air
uuuu
1_
§ 5000.
c
o
'(^ A AHA
•§ 4UOO
o
La
o- 3000
V)
S 2000
>
43
tt
1000
0
-
"•
-

I TA100







I \ Gas Phase
Paniculate Phase


TA98

          BEFORE
        IRRADIATION
             AFTER
            IRRADIATION
  BEFORE      AFTER
IRRADIATION  IRRADIATION
  Figure 2. The effects of irradiation on the gas phase and particulatc phase mutagcnicity of diluted auto exhaust,
       using two different strains of bacteria.
                            889
-------

FINAL DESIGN AND FIELD  EVALUATION  OF
THE HIGH VOLUME PM2.5 VIRTUAL  IMPACTOH
Robert M. Burton & Alan J. Hoffman
Monitoring and Assessment Division                        «/
Environmental Protection Agency
Research Triangle Park, NC 27711                           \

Virgil A. Marple, Ph.D.                                        "^CHi
U. of Minn. , Particle Technology Laboratory
Minneapolis, MN
                                                                          ^
     Final design and  laboratory testing  have  been  completed for the •'
High Volume PM2.
-------
 IQ   The  purpose  of the  High Volume  PM 2.5  Virtual Impactor Design Task was
   provide  the  Integrated Air Cancer Project  (IACP) with  a  particulate sam-
     capable  of separating  and collecting large amounts of  PM10 particulate
 jter in two distinct fractions  (0  - 2.5 microns and 2-5  - 10 microns ).*
 ^sampler's final design  incorporated twelve parallel  "paired" opposing
 •lov  1 impactors  operating at a total flowrate  of  kO  CFM.2   The  coarse
   v  was  limited to  5% of the total  flow (2 CFM).   Since both fractions of
     Hlate matter are collected by filtration, greasing the collection sur-
   t   not re1uired (as  for cascade  impactors), thus eliminating the possi-
   L  °f samPle contamination.  With the sampler operating  for 12 hours at
   40 CFM flowrate (38 CFM  for fine fraction), sufficient  amounts of organic
        to be used for bioassay are collected.
            METHODS
•V ^° reduce the high pressure buildup in the sampler, the earlier design
S$(je incorporated six  parallel paired virtual  impactors  had to be  super-
W  by twelve  parallel  paired impactors.   The new  design lowered  the
    nal pressure dr°P  and facilitated  field  sampling  by requiring  less
    capacity  for movement of  sample air  through  the impactor assembly.
       twelve  virtual  impactors  were  calibrated  using monodispersed oleic
             and ammonium fluorescene  (solid) particles  generated with a
        orifice  monodispersed aerosol generator (VOMAG).3  Minor adjust-
SD? *n iraPactor  geometry were made to  minimize particle  losses inside the


Vrs
        e 1 provides  both  calibration (collection efficiency vs particle
    and particle  loss  (by  size)  for  the final  configured  sampler.   The
     lection efficiency occurs at 2.5 microns for liquid particles and at
     mately 2.3 microns for the solid calibration particles.
            loss  is  greatest  for  liquid particles.   As  noted in  the
      liquid particle loss  reaches  a maximum at the cutpoint.   Loss,  as
     ned by solid particles,  is  shown to  be less than 2% below the  2.5
 .    outpoint.   In most cases, solid test  aerosol  resembles more closely
 ,icnaracteristics  of ambient particulate matter, or closer than the liquid
 lge acid particles.   The coarse mode filter has  been designed  to lay very
  r^.0 the  exit  of the  receiver tubes (1.6 inches),  minimizing the chance
     rse mode particles to be attracted to internal walls of  the sampler
     being  collected  on  the  filter.

 >(Whe  aamPler utilizes  the standard 8"  x 10" hi vol  filter  and  filter
 *«i-  f°r the fine  fraction particle  collection, and  a 2"  x 6"  filter with
  i   holder for the coarse mode collection.   The filters  can be  loaded  in
   moratory  for  ease  of operation.   A  cover  for the  filter  cassette
   'ards the  filter  assembly  for transport  between the laboratory and
   site.
              °f tne  impactor body  is performed  by casting  to insure
      Jnensions and tolerances  of critical  components  for repeatability
  r °rmance.   For  calibrating and  setting  the flowrate,  a top loading
  i     adapter has been built to allow use of the standard top loading
    calibration/audit orifice assembly.
                                 891
-------
     Most recently,  the new sampler is being  tested in a field  evaluati°
sampling study at Durham, NC, to test its  field adaptability  and         '
ity to  the  standard  l6.7 Jl/min  dichotomous   sampler.1*  5    The
though more complex than the hi vol TSP sampler, functions well as a
particulate sampler.  Filters seal well and  flows  remain constant at  38
and 2  CFM,  respectively,  for  the fine and coarse  sample  fractions.
impactor assembly fits  in both the Andersen  and Wedding PM^Q  Size
Inlet hi vol samplers.  Figure 2 illustrates the comparability of fine
collected simultaneously  by the new impactor  and the standard  l6.7
dichotomous sampler  during  the  first  five days of  sampling at  the  Dur'1 1
site.  The fine  masses  for the two  samplers had a  correlation coeffic16
of .99 with the  new  impactor mass approximately 9% higher than mass for *
standard dichotomous sampler.
     An induction motor coupled with an electronic  volumetric  flow
ler is  in final design  and testing  to provide ease  of operation,
reliability, and flexibility in flow measurement and control,  The  indue
motor, converted electronically to three-phase, offers the  following a<*v
tages :

     1.   No brushes to wear out, therefore no periodic  maintenance.

     2.   Ho brushes to contaminate the sample air.

     3.   Easier to control the speed, therefore higher  accuracy.

     U.   Much quieter operation

     5.   Approximately 20% less energy use at 1*0 CFM.

     6.   2-3 times longer blower life since it operates at  10,000 rpfl
          instead of 20,000 rpm.

CONCLUSIONS
                                                                       ^
     The High Volume PM2.5 Virtual Impactor operates as designed  for an1flfj'
on to either Andersen or Wedding PM]_Q Size Selective Inlet hi  vol  samp* ^
If the upper cut of PM^o  *s not  required,  the  impactor can be operated * ^
pendent ly in a  high volume  shelter.   The  sampler will be cost-effe  ^
since it utilizes the standard high volume filter,  filter holder,  and
                                                                       *0 CFM sample provides sufficient particulate loading for
testing.
                                   892
-------
REFERENCES

l*    U.S.  EPA Clean Air Act,  Section 112.

2«    R. M.  Burton,  V.  A.  Marple,  and B.Y.H.  Liu,  and B.  Johnson,  "A novel
     2.5 ^m  cut  virtual impactor for high  volume PM10  sampling,"  Proceed-
     ings  of  the 1987  EPA/APCA Symposium on Measurement of Toxic and Related
     Air Pollutants, Research Triangle Park, NC  (May 1987).

^'    R. N.  Berglund and B.Y.H. Liu,  "Generation of monodisperse aerosol
     standards," Env.  Sci. &  Tech. J:lU7-153 (1973).
I,
     B. W.  Loo, J.  M.  Jaklevic, and  F. S. Goulding,  "Dichotomous  virtual
     impactors for  large  scale monitoring of airborned particulate  matter,"
     Lawrence Berkeley Laboratory Report No.  LBL-3851* (1975)*

*   T. G. Dzubay and  R.  K. Stevens,  "Ambient air analysis with the
    dichotomous sampler  and  x-ray fluorescence spectrometer,"  jSnvir. Sci.
    _and Tech. £, 7:663(1975).
                                 893
-------
00
CD
                                          23                5          7
                                  Aerodynamic  Particle Diameter  (Microns)
                                                                                                  Liquid Particle
                                                                                                    Calibration
                                                                                                 Liquid Particle
                                                                                                      Losses
                                                                                                  Solid Particle
                                                                                                    Calibration
Solid Particle
   Losses
                           \«  Si_z.e sepa.rati.oTx characteristics for High Volume ^^2.5 Virtual. Impactor.
-------
                    Slope = 0.9236
                    Correlation Coefficient
                    N =  5
                    Intercept =  -0.6613
CO
                           30     40     50    60    70     80    90

                             HIGH VOLUME PM  2.5 FINE  MASS. U6/M3

            Figure 2.  Fine fraction particle mass comparison  (0-2.5 microns).
100    110   120
-------
QUALITY ASSURANCE PLAN USED AT THE LOVE CANAL
EMERGENCY DECLARATION AREA INDOOR AIR
ANALYSES BY A TAGA 6000E MASS SPECTROMETER /
MASS SPECTROMETER
Thomas H. Pritchett
U.S. EPA Environmental Response Team
Edison, New Jersey

David Mickunas & Nickolas Kurlick
Roy F. Weston Inc.  (REAC)
Edison, New Jersey
     A quality assurance / quality control plan was develop6^
for the TAGA 6000E mobile mass spectrometer / mass spectromet
which enabled the instrument to be used in the Love Canal
Emergency Declaration Area habitability study.  Because of
described QA/QC the TAGA was able to produce reliable,
defensible data.
     Eight data quality objectives were defined for the
seven of which were directly related to the instrument or
data.  These objectives set criteria for accuracy, precisio*1'
detection limits, sensitivity decay, calibration accuracy,
sampling efficiency and completeness of data documentation.
Several procedures were established to ensure that these
objectives were met.  In addition, several potential sources
for errors in TAGA analyses were identified.  The paper
describes quality control actions taken to minimize the
occurrence of such errors.  Finally, several sets of summary
QA/QC are presented to illustrate the success of the plan.
                             896
-------
introduction

     In the Autumn of 1986  the U.S.  EPA Region II  made a
        to the Environmental Response Team (ERT)  for TAGA
    -_  Atmospheric Gas Analyzer 6000E Mass Spectrometer / Mass
Spectrometer)  indoor air analyses in support of the Love Canal
Emergency Declaration Area  Habitability Study.   The final, full
Sampling program was to be  completed and a final report
9enerated within the first  two months of 1988.   This report,
arifl any requested raw data  packages, would then undergo an
Eternal review by a peer review committee of individual
Scientists not associated with any agency or contractor invol-
jed with the various Love Canal projects and decisions.   Final-
iir, the generated air data,  in conjunction with data from
c°ncurrent soil and dioxin  sampling  programs, would then be
Used by the commissioner of  the New  York Department of Health
j° decide the  future habitability of the homes contained in the
lj°ve Canal Emergency Declaration Area (EDA).

     The Love  Canal Technical Review Committee (TRC),  which is
Composed of representatives  from the state and federal agencies
 Evolved with  Love Canal, selected a set of indicator  compounds
'kClCs)  for which all air analyses would be performed  .   The
fleeted compounds had to meet the following criteria:  1)  had
     disposed  of at the canal,  2)  were not ubiquitous  to the
       Falls area, 3)  had been shown in past data  to correlate
     levels of other compounds present in the Love Canal
     s,  5)  could be easily  analyzed  for in air,  and 6)  had to
   covered by  a relevant state or federal criteria or  guide-
  -es.   Three  compounds were thus selected:  2-chlorotoluene,  4-
^lorotoluene,  and chlorobenzene.  Because of the  TAGA's
.^ability to distinguish between the three chlorotoluenes  which
fave the chlorine attached to the aromatic ring, this  list was
 u^ther  compressed to just chlorobenzene and chlorotoluenes.

     In  1986 a pilot study was  conducted to evaluate the feasi-
       of using the TAGA for the air portion of  the habitabil-
    study.   The results of the pilot study2 were also  used to
    ne  the specific needs to be addressed in the Project Qual-
    Assurance  Project Program (QAPP)  in general  and in the TAGA
        assurance plan specifically.   These results and recom-
           were publicly commented on by the Love  Canal Peer
      Committee during a public meeting of the  TRC in late
         of 1987.   These comments  and the TRC replies  were then
            into the developed sampling plan,  the overall
     and the  TAGA quality assurance plan.

^t  A key finding from the  pilot  study was the extremely  low rate
ha w^^cn  tne target compounds were detected in the sampled
 uines from both the EDA (1/63)  and the control areas  (1/31).
;*,E!?cl upon this data,  no meaningful  statistical conclusions
     be  drawn  unless  an excessive number of  control homes were
          This  finding resulted in the TRC deciding not  to
        data gathered from the  EDA homes and  the comparison
    homes.  Instead,  the presence or absence of any LCIC  in
                             897
-------
the indoor air house became the preliminary indication of al
habitability of that house.  Whenever an LCIC was detected
within a house, an attempt would then be made to determine ^
emission source.  The revised sampling plan required that
structurally sound residential building had to be sampled
during one of the four phases of the study (provided that
homeowner permission could be obtained) and that the samp
schedule had to be sufficiently randomized to avoid potenti3
biases due to seasonal, diurnal, and other variations.

     The overall project management was the responsibility ° &
CH2M Hill, Inc., the Love Canal Remedial contractor, under *
direction of U.S. EPA Region 2.  CH2M Hill assumed responsi
bility for designing the overall sampling program to i
scheduling, responsibility for generating the overall
Assurance Project Plan1, and responsibility for coordina
the review and sampling processes.  The ERT assumed
bility for developing and refining the required TAGA
methodologies, developing the quality assurance plan for
TAGA generated data, and for writing the "Summary of the
Standard Operating and Reporting Procedures for the Love
Full-scale Air Sampling study", (SORP) which was incorporate
as Appendix A of the QAPP3.  Northrop Services, the
contractor for the Quality Assurance Branch of the EPA's
ronmental Measurements Support Laboratory at Research Tr
Park, North Carolina (EMSL/RTP), agreed to provide external
field QA audits and TAGA performance evaluation (PE)
during all phases of the study.  The TRC had final approv
authority over all phases of the sampling program
overall and TAGA specific quality assurance plans.

     The designs and approvals of sampling plan and QAPP
also heavily influenced by the past scrutiny and severe
cal reviews  previous studies at Love Canal had undergone-  ^
Consequently, the full QAPP, which incorporated both the TA
SORP and the sampling program, was not approved until the
document had been thoroughly reviewed by all agencies repr
sented on the TRC and by the Quality Assurance Branch of
EMSL/RTP.

Procedures

     Appendix A of the "Love Canal Full-Scale Air
Study Quality Assurance Project Plan"3 contains a
discus sion of the standard operating and reporting
used by the TAGA group during this study.  The procedures
directly related to the TAGA Quality Assurance plan are ^^
rized below.

                   Initial Instrument Tuning

     Each day the townsend discharge current was set to 1°J
microamps and the source pressure was adjusted to .9 to !•
torr.  The high voltage quadrupole power supplies were tn
allowed to warm up for at least thirty minutes.  A tetra-
                             898
-------
chioroethylene and trichloroethylene vapor mixture was then
introduced to the sampled ambient air stream.  After the elec-
tron multiplier voltage has been optimized, the 85+, 130+, and
166 + parent ions were scanned for on each quadrupole.  During
the scanning the ion intensities are optimized by adjustments
to the rod off-sets while the quadrupole resolutions are set to
Yield essentially equivalent mass peaks with widths at half
height in the range of 0.55 - 0.85 AMU and with reasonable mass
Peak shapes (i.e., no major splitting of peaks and semi-rounded
Peak tops).  This was an iterative process since changes in the
rod off-sets change the peak resolutions and peak shapes while
changes in the peak widths affected the observed ion intensi-
ties.   After an optimum set of quadrupole parameters had been
found then a mass calibration was performed on the quadrupole.
°nce both quadrupoles had been tuned then the instrument was
ready for compound calibrations using the procedures below.

                     Instrument Calibration

     TAGA calibrations were performed by diluting primary gas
fixture standards (Scott Specialty Gases; Plumsteadville, PA)
lrito ambient air being pulled past the TAGA inlet using the
shipment shown in Figure 1.   The primary gas standards typi-
caiiy contained from 20 - 30  ppm of each component in dry
Nitrogen.   The standard gas flow was regulated by a 0 - 100
^l/min mass flow controller (MFC).   The ambient air flowrate
vas not only controlled by a  manual adjustment valve but the
flow was continuously measured by the TAGA with a Halpern
transducer which was downstream of the TAGA inlet.   This flow,
vhich the TAGA refers to as the SAF, typically ranged from 1350
" 1500 ml/sec.

      Measurement of Sampling Line Transport Efficiencies

     Twice daily the transport efficiency of the sampling
    s, which were 100 - 200 feet in length,  were determined for
    LCICs  using the equipment shown in Figure 2.   The first
J-°upie of  feet  of the sampling lines were fed back into the
:Jus,  but the bulk of the lines remained outside exposed to
I^ient temperatures.   Flow from a primary gas standard was
^Uvered through the dilution system at MFC and SAF flows
^ich  would yield a diluted concentration of 3 - 6  ppb of each
 opponent.   The three way valve was initially set so that the
^tandard gas by-passed the sampling line.   After the signals
:°r both LCICs  had equilibrated, data was averaged for at least
 5 seconds (to  insure that at 30 individual  measurements were
         .   The three way valve was then turned such that the
        was introduced at the head of the sample lines and the
    was recorded between the initial drop in the LCIC signals
   their return back to the previous levels.   This  measured the
 esident time of the sample in the  lines which was  then used to
        future  house data prior to  reduction.   Once the signals
    restabilized another 45 seconds of data  were averaged.   The
 ^nsport  efficiency was calculated from the upstream average
 I9nals (Su)  and the initial  downstream average (Sd)  by
                             899
-------
     % Transport Efficiency = (Su / Sd) * 100.

     During the first phase a heated transfer line, which
always remained at least 70°F, was used.  After experimental
work between phases 1 and 2 demonstrated that no sample loss
would occur at 32°F, unheated lines were used for the remain!119
phases.  However, during phases 2-4 the TAGA operators were
required to always record the ambient temperature for every
transport efficiency experiment and for every house analysis-
The lack of sample loss at lower temperatures was further
confirmed by the transport efficiency data from the subsequent
phases (Figure 3).

                     QA/QC Sample Analyses

     During the study the TAGA group was routinely required
analyze both internal and external quality assurance and
ity control sample in order to measure the precision and
racy of the instrument.  All of these analyses used the same
procedure and differed only in the type of sample being
analyzed and in how the data was processed.  In all cases
sample was introduced to the TAGA using the same dilution
system used for performing the calibrations (Figure 1).  The
diluted concentrations typically ranged from 5-15 ppb.  The
samples used included the Scott primary standards (internal
accuracy checks and precision), 16 Liter Summa canisters (eX-^
ternal accuracy checks and precision) and 6 Liter Summa canis"
ters (external accuracy checks).  The measured errors were
calculated by

     %Error = [(CQ - Cc) / CQ] * 100

where CQ was the original concentration (ppm) in the sample ^
Cc was the undiluted sample concentration (ppm) derived from  .
the dilution data and the concentration (Cm in ppb) reported
the TAGA.   cc, in turn, was derived from

     Cc = (CQ * SAF * 60) / (MFC * 1000)

where SAF is the measured ambient air sample flow (ml/sec)  afl
MFC is the flow (ml/min) at which the sample was introduced
into the ambient air.

Results and Discussion

     Several goals were defined for the TAGA's quality
assurance plan.  These goals were:

1) to ensure that the TAGA had the required sensitivity for ^
     analyses;
2) to have sufficient quality control procedures to ensure
     reliability and legal defensibility of all TAGA data
     generated;
                             900
-------
3) to incorporate procedures for ensuring that the various key
     components of the sampling equipment were performing
     within the required specifications and for ensuring that
     the TAGA was acquiring data with the required accuracy and
     precision;
4) to incorporate measures to monitor the drift in the TAGA's
     sensitivity and to correct for response factor drift
     whenever such drifts occurred;
5) to ensure that, for all sampled residences, the sampling,
     data acquisition, and data reduction documentation were
     all consistent with each other and were complete;
and 6) to ensure that the above five goals could be
     accomplished quickly enough such that no additional
     mobilizations would be required for resampling homes with
     previous data of unacceptable quality.

     Table I summarizes the TAGA data quality objectives (DQOs)
which were derived from these goals.   Three different criteria
levels were used for these objectives.   The most severe crite-
ria (R), although associated with the documentation rather than
the instrument performance, required that additional (or
replacement) sampling be performed - even if an additional
Mobilization would be required.  If the instrument performance
vas substandard for a given house (e.g., detection limit was
accidentally above 4 ppb), the house package would have been
been rejected as incomplete because an acceptable detection
Umit could not have been documented.   The other two criteria -
both of which were used for objectives related to instrument
Performance - required corrective actions to be performed prior
to additional sampling.

     The least restrictive "G" criteria, which was only applied
to the accuracy objective, allowed sampling to restart even
vhen it could not be accurately determined that this objective
vas being met based upon the established criteria.  The "G"
criteria was initially applied to the accuracy criteria "as
Determined by performance evaluation analyses" because of the
experimental nature of using the 6 Liter Summa canisters as
TAGA performance samples.  Thus, when problems were initially
^countered with the use and certification of the 6 Liter Summa
pE canisters, alternate means for directly and indirectly mea-
suring the TAGA's could be used and the first two phases of the
study did not have to be rescheduled.

     Several procedures, both internal  and external, were used
to ensure that the quantitative accuracy of the instrument met
the data quality objective of a |%error|  <. 25%.   First, during
e^ch phase of the study the Northrop Services QA auditors
challenged the TAGA with a blind performance evaluation (PE) 16
^iter Summa canister sample.   Prior to  its use the canister
Concentration had been certified within 10% of the theoretical
concentration.   Secondly, Northrop Services also was tasked to
SuPply the TAGA group with daily blind  PE 6 Liter Summa canis-
ters.   Unfortunately, problems with the Northrop certification
         prevented the 6 Liter Summa's  from being used until
                             901
-------
the last two phases.  Third, during each phase Northrop
Services also supplied us with two 16 Liter canisters of known
concentration for use in determining the precision of the TAGA-
However, starting in phase 2 these samples were also used as
internal control samples.  Fourth, at the start of every day
the TAGA group also analyzed one of the Scott standard gas
cylinders as another internal control sample.  The analyzed
cylinder was never the same cylinder with which the instrument
was calibrated.  Finally, at the end of each day the TAGA <3*°,
spiked the ambient air with both LCICs at concentrations with11
1-2 ppb of the reported quantitation limits in order to
document the quantitative accuracy of the instrument at the
quantitation limits.

     The results from the first four types of accuracy checks
are summarized in Table II.  The measured error in these ana-
lyses exceeded the 25% criteria only once during the whole    ^
program and on that day only one of three separate analyses n<*
an excessive error.  The 25% error criteria was never exceeded
during any of the quantitation limit verification spike
analyses.

     The results from the third and fourth type of accuracy
checks were also used to measure the within phase precision
the instrument.  In all cases the within phase standard
tions from the mean were less than 25%.
     The third criteria set a maximum allowable detection
of 4 ppb for either LCIC.  Table IV contains a summary of the
detection limit data for all four phases of the study.  Only
once did a reported detection limit exceed this criteria and
then its value was only 4.2 ppb.  Even though this value woul
have rounded down to 4, the applicable house was selected fo3r-f
replicate sampling during that phase and all subsequent phase*"
     The fourth criteria set a maximum for the allowable
 (15%) in TAGA response factors.  This criteria was designed
only to catch decreases in the TAGA's sensitivity.  The TAGA
plan had built into it procedures to correct for the effect
sensitivity changes a given set of quantitations and to com  ^
the resulting potential error in the reported value.  In add1 ,
tion, preliminary work had shown that the TAGA's response wo1* 4
vary throughout the day and had a tendency to increase.  Dui*1
the study  it was not unusual to find a doubling of response
factors within the first several hours of instrument operati0 f
However, because of the emphasis on determining the presence
absence of LCICs and because of the procedures developed for
correcting data for drifts in instrument response, the prima"
reasons for monitoring changes in the instrument's sensi-i-ivi^
became insuring 1) that the instrument detection limits
never exceed the maximum acceptable detection limits and 2)
that no false negatives would be encountered because of a »"^-^
drop in instrument  response  from the time of pre-entry calif*
tions and the actual analyses of the houses.  In every case
that an unacceptable decay occurred the appropriate correctiv
                              902
-------
 Actions were taken prior to the sampling continuing.

                Other QA/QC Procedures & Results

      Prior to defining the various quality control procedures
  o be used, we first identified the various potential sources
  * error.   These sources were 1} drifts in instrument response,
  J errors  in the actual calibrations, 3) loss of the LCICs to
  ne sampling line walls during transport of the air from the
 Bouses to  the TAGA,  and 4)  manual data entry errors.

      Gradual drifts  in the  instrument sensitivity could arise
  J-om such  uncontrollable causes as changes through the day in
  Jje sample matrix (e.g., relative humidity)  and in variations
  Jj the ion transport from the source through the quadrupoles to
  Jje. detector which could be caused,  for example, by gradual
  ^cumulations of electrostatic charges on the various focusing
  enses.  To correct  for these errors two quality control proce-
  ures  were utilized.   First,  new calibrations with new detec-
  jon limits were required for each analyzed  house.   In addi-
  *on,  the  decay  in response factors  between  successive calibra-
  lon was monitored to  insure  that  no drastic  drops in sensitiv-
  ty  were occurring.  Second,  if  an LCIC  was  detected within  a
  °use  and  an investigative  survey  was  initiated,  then a subse-
 jjient  calibration  was performed  prior  to leaving the house.
 g£e  reported concentrations would  then be determined by using a
 st  of intermediate, or  inverse  average,  response  factors  (IRF)
 srived from the initial  a    eliminated by the  requirement  that  all key instrument
6hfail'eters be documented  for each analysis by either  manual
tK   ,es on various logsheets  (source  pressure)  or hardcopies of
    ^nstrument  parameter  table.  For  each analysis  these para-
 ^6r entries were then checked  by  both  a data review group  and
 0ri  a validation  group to  insure either that they  remained
 astant  (e.g., quadrupole  settings,  electron multiplier
                            903
-------
voltage, etc.) or that they were within a previously estab-
lished operating range (e.g., collision gas thickness, source
pressure, etc.)

     The errors in the calibrations could have arisen from
either incorrect values for the primary gas standard concent*"3
tions or from  errors in the dilution flow measurements.  The
calibrations of the SAF measurement transducer and of the    ,
standard gas mass flow controller were checked at the start ol
every phase and every seven days thereafter.  In addition,   e
Northrop Services, Inc. performed an external flow audit on tP
SAF and MFC at least once every phase.  These audits were
performed using laminar flow elements with NBS-traceable
brations.  To minimize the error due to incorrect values for
the primary gas standards, a program of periodic recalibrati0^
analyses was established for these standards.  Each recalit>ra
tion usually consisted of separate analyses by both Scott
Specialty Gasses and Northrop Services.  The total number of
certification / recertification analyses for each cylinder
ranged from two to four analyses at each laboratory over a
seven month time period.

     Very early in the QA/QC plan design we became concerned ^
about the potential for loss of low levels of LCICs to wall5
the sampling lines.  We established two QC checks to monitof
for this possibility.  First, at the start and end of each
sampling day the TAGA crew was required to document the tr^1 ^
port efficiency of >. 85% for the sampling lines external to
mobile laboratory.  Until experimental work between phases ^ &
and 2 documented that no LCIC would be seen at temperatures
low as 32°F, a heated 200' teflon 7/8" I.D. hose was used t^
pull the air from the house to the TAGA.  This hose was mai*1
tained at 70°F when used in sampling.  During phases 3 - 4  ^
unheated 7/8" Teflon hoses were used but the ambient tempefa
ture was recorded.  This temperature was always compared
against the lowest temperature for which we had transport
efficiency data.

     Figure 3 contains a summary of the transport efficiency
versus temperature results from the last three phases.  NO
correlation between transport efficiency and temperature
found for either LCIC.  With the heated transfer line, the
transport efficiency for chlorotoluene and chlorobenzene
averaged 100.2+8.5% and 100.8±5.8%, respectively, with m
measured efficiencies of 90.4% and 90.9%.  With the unneate
transfer lines the transport efficiencies for chlorotoluene
chlorobenzene were 97.7±5.3% and 98.5+4.2% with minimums of
88.6% and 91.4%, respectively.

     The second check was designed to ensure that trace l^^
of the LCICs would not be lost to the TAGA inlet lines the*1
selves.  On a daily basis the TAGA group spiked the ambie11.
stream at a concentration less than 1 ppb above the detect*
limit and verified that the TAGA would indeed detect both
LCICs.  These two repetitive quality control analyses
                             904
-------
 essential to the defensibility of the air data because of the
 very high number of "not-detected" determinations which were
 reported.

      Several procedures were established to avoid,  catch,  and
 easily correct the expected manual data transcription  errors.
 With projected long working hours and the shifting  of  sampling
 start times, crew fatigue and the associated data entry errors
 vere anticipated for each phase.   Therefore,  redundancy was
 built into the design of the house data packages.   Although
 each form was designed for a specific purpose (e.g. , data
 Acquisition flag number versus time,  house floor  plan  and  the
 hose sampling location versus time and data flag) ,  all  manually
 Qntered data had three different  forms on which it  was  to  be
 transcribed and the transcription responsibilities  resided with
 at  least two individuals.   This redundancy allowed  the  vast
 Majority of found data entry errors to be corrected without
 Caving  to either revisit the site or  question the sampling crew
     Internal data  review  and external data validation groups
 *ere designated and given  responsibility to catch the manual
 transcription errors, other documentation errors {e.g., a
 specif ic form missing from house package) , unapproved devia-
 tions from the QAPP, and failures of the data to meet the
 specified data quality objectives.  Because of the DQO of 100%
 Complete and consistent documentation for every house sampled,
 the DQO of 100% sampling rate for the designated residences,
    the high costs  for remobilizing the full sampling team, on-
     review and validation was utilized to ensure that all data
     ges could be finalized within 48 hours by the last mobili-
 ation.   Both on-site groups were set up at the motel during
 ach phase of sampling.   Each group had predefined checklists
 to follow (Tables 7.1 & 7.2 of Appendix A of the QAPP3).

 w    The data review group, which was staffed by the TAGA crew,
 as responsible for ensuring that each documentation package
 as complete, that all data entries were consistent throughout
    forms for that house,  and that all manual data entry errors
     indeed corrected.   After a data package was reviewed it
    forwarded onto a group from the ERT's Technical Assistance
     (TAT)  contractor for  data validation.

     The data validation group was responsible for  finding any
    inin  data inconsistencies and errors,  for documenting on a
         logsheet  to the ERT QA/QC coordinator the  results from
    various  DQO  related  QC  analyses,  and for flagging any
 ^tential  problems which might affect the data quality of a
v Ven data package.   When  such a problem was found,  the problem
 ,s described on part  2  of  the data validation comments log-
     '   Tne  ERT  QA/QC  coordinator would  then review  the problem
     and then decide to  which  category the problem belonged:

      "problem" does not affect  the  data quality -  no action
     required;
                             905
-------
2) a documentation problem does affect the data quality but J
     can be corrected by the data review group;
3) the data quality has been affected according to a strict
     interpretation of the QAPP and an explanatory memorandum
     to file is needed from the ERT QA/QC coordinator to
     correct the problem;
and 4) the problem is not correctable - the house must be
     resampled.
     All of these determinations could only be made by the
QA/QC coordinator and his decisions were recorded on the a
cable comments logsheet.  In several cases, memorandum to
were needed to modify the QAPP procedures when various uneX'
pected problems occurred in the field.  For example, during
phase 1 one such memo modified the detection limit criteria
greatly decrease the probability of high instrument noise
causing false positive readings.

Conclusions

     Although the TAGA had had a previous reputation for
questionable quantitative accuracy, the technology proved
it could produce reliable, defensible data provided that
proper QA/QC procedures were used.  Even though the frequency
and redundancy of quality control procedures were higher ^ .
normal, many of the procedures developed for this program &^
applicable to future TAGA projects and have been incorporate
into the ERT's evolving TAGA QA/QC plan.

Acknowledgements

     We would like to acknowledge the assistance of CH2M
in the preparation of the TAGA standard Operating and Repo
Procedures for this project.  In particular Gary Helms and
Barbara Hart were invaluable with their suggestions and '
their overall help during the preparation of the written
plan.

References

1.   "Love Canal Emergency Declaration Area; Proposed
     Habitability Criteria,"  New York State Department of
     Health, Albany, NY, 1986.

2.   "Pilot Study for the Love Canal EDA Habitability
     CH2M Hill Southeast, Inc., Reston, VA 1987.

3.   "Love Canal Full-Scale Air Sampling Study Quality
     Assurance Project Plan,"  CH2M Hill Southeast, Inc./
     Reston, VA, 1987.

4.   "Habitability of the Love Canal Area, An Analysis of ^
     Technical Basis for the Decision on the Habitability °
     the Emergency Declaration Area,"  Office of Technology
     Assessment, Washington, DC, 1983.
                             906
-------
     I.   TAGA Data Quality Objectives for  the Love Canal EDA
         Air Habitability Study.
                  Objective
1.



2.



3.

4.


5.
6.
Overall TAGA Accuracy as Determined
by Performance Evaluation Analyses
(% error of reported concentration)

Overall TAGA Precision as Determined
by Periodic Analyses of Same Cylinder
{% error of reported concentration)

Detection Limits for LCICs (ppb)

Allowable Decay Between Consecutive
Calibrations (% decay of ion signal)

Accuracy of TAGA Calibration
(% difference, reported vs. actual)
  Sample Air Plow
  Mass Plow Controller

Total Loss of LCIC in Sampling Lines
(% loss of ion signal)

Residential Structures Sampled
(% of EDA structures for which
permission to gain entry was granted
and for which the structure was
determined safe to enter)

Documentation Complete and Consistent
for Sampled Residence (% complete)
Criteria

  £25%



  325%



   £4

  £15%
                                                          Criteria
                                                              Key
C

C
  £10%
  £10%

  £15%
                                                    100%
                                                                   C
                                                                   C
                                                    100%
(S
   Goal  that, when not met, requires corrective action.  If the goal is
   not met  after  the corrective action, additional performance
   evaluation analyses will be performed.  If the errors are still >25%
   but are  consistent  (error range of less than 10%),  then the error
   "ill  be  considered a systematic bias.  This bias will be noted,
   efforts  will be initiated to determine the cause of the bias, and the
   sampling restarted.
Q
 ""Absolute criteria that must be met before continuing the job.
K
 ""Documentation  goal that will require corrective actions to include
   Additional sampling, if necessary.
                                907
-------
TABLE II.  SUMMARY OF THE OVERALL ERROR RESULTS AS MEASURED BY VARIOUS PE ANALYSES
1
1
1
1
1— -~^^
1 PHASE 1 DATA
1
I AVERAGE PERCENT ERROR
|AVG MAGNITUDE OF % ERROR
1 %ERROR w/ MAX MAGNITUDE
1 PHASE 2 DATA
1
I AVERAGE PERCENT ERROR
|AVG MAGNITUDE OF % ERROR
t %ERROR w/ MAX MAGNITUDE
1 	 	 _
1
I PHASE 3 DATA
1
1 AVERAGE PERCENT ERROR
|AVG MAGNITUDE OF % ERROR
1 %ERROR w/ MAX MAGNITUDE
i
1 	
1 PHASE 4 DATA
1
I AVERAGE PERCENT ERROR
|AVG MAGNITUDE OF % ERROR
I %ERROR w/ MAX MAGNITUDE
1 AS MEASURED
BY
1 SCOTT CYLINDERS
1
1 CHLOROTOLUENE CHLOROBENZENE
1
2.2%
I 4.9%
23.0%
	
-0.3*
5.4*
11.5%


-6.6%
13.7%
-22.4%


-0.4%
8.4%
-22.2%

-1.7%
5.6%
15.6%

-2.9%
5.9*
-9.7%


-10.9*
14.9*
-23.1*


-4.8%
6.6%
-20.7%
1 AS MEASURED BY |
1 6L SUMMA CANISTERS |
1 i
I CHLOROTOLUENE CHLOROBENZENE |
1 I
1 1
1 ERROR DATA NOT COMPUTED |
I FOR THIS SET OF DATA I
I (REFER TO TEXT FOR REASONS |
t 1
| ERROR DATA NOT COMPUTED |
I FOR THIS SET OF DATA |
I (REFER TO TEXT FOR REASONS |
1 1
| |
1 -8.7% -9.1% I
1 10.4% 9.5% |
1 -24.3% -25.0% |
1 1
1 |
1 -4.1% -1.9% |
I 13.4% 10.0% |
I -18.8* -22.0% |
AS MEASURED BY
16L SUMMA CANISTERS
CHLOROTOLUENE CHLOROBENZENE

ERROR DATA NOT COMPUTED
FOR THIS SET OF DATA
(REFER TO TEXT FOR REASONS

2.0% 'I*
3.4% 3'6*
5.3% ~7'4*


-16.6* ~13'
16.6* U 5*
-26.4% ~24


-10.6% '4-'*
4 4*
10.6*
-22.1% ~9' „.

                                           908
-------
TABLE III.  OVERALL PRECISION RESULTS AS MEASURED BY VARIOUS PE ANALYSES
1 1
1 1
1 1
1 1
1 	 _ 	 	
1
| PHASE 1 |
1 CYL ID E, # MEASUREMENTS I
| AVG CONCENTRATION (PPM) |
| RELATIVE STD DEV, |
II
1
I PHASE 2 I
| CYL ID & * MEASUREMENTS |
| AVG CONCENTRATION (PPM) |
| RELATIVE STD DEV. [
i i
1 — 	 1
I PHASE 3 |
| CYL ID & It MEASUREMENTS |
| AVG CONCENTRATION (PPM) (
| RELATIVE STD DEV. |
1 1
| CYL ID & tt MEASUREMENTS |
| AVG CONCENTRATION (PPM) I
I RELATIVE STD DEV. |
Ij
	 	 j.
I PHASE 4 )
| CYL ID & * MEASUREMENTS |
| AVG CONCENTRATION (PPM) |
I RELATIVE STD DEV. |
Ii
	 1
| SUMMARY |
1 AVG RELATIVE STD DEV. I
I MAX RELATIVE STD DEV. |
AS MEASURED BY
j AS MEASURED BY
SCOTT STANDARD CYLINDERS )

1
CHLOROTOLUENE CHLOROBENZENE |

AAL-9226
26.1
3.9%

AAL-15177
26.1
6.5%

AAL- 15246
29.7
5.3%

AAL-9562
28.4
17.9%

AAL-16555
30,1
10.6%



8.8%
17.9%
1
(10) |
26.9 |
4.8% |
1
(11) 1
26.3 |
5.5% |
1
(5) I
28.7 |
2.4% |
1
(5) |
28.4 |
15.6% |
i
	 — 1_
1
(10) I
28.6 |
7.2% |
-~— ~— i _
1
1
7.1% |
15.6% |
16 LITER SUMMA

CANISTERS

CHLOROBENZENE CHLOROTOLUENE
Two 16L canisters each
analyzed four
Therefore, no
data reported,

PEB-lA
37.9
3,3%

PEB-1B
46.1
5.5%





PEB-1C
38.3
2.7%



3.8%
5.5%
times.
precision


(5)
30.8
3.1%

(11)
40.7
5.4%





(4)
34.4
2.6%



3.7%
5.4%
                                      909
-------
TABLE IV.  SUMMARY OF DETECTION LIMITS  USED FOR HOUSE
          ANALYSES DURING THE LOVE CANAL  EDA FULL AIR STUDY

PHASE 1
AVERAGE DETECTION LIMIT (ppb)
STANDARD DEVIATION (ppb)
MINIMUM DETECTION LIMIT
MAXIMUM DETECTION LIMIT
PHASE 2
AVERAGE DETECTION LIMIT (ppb)
STANDARD DEVIATION (ppb)
MINIMUM DETECTION LIMIT
MAXIMUM DETECTION LIMIT
PHASE 3
AVERAGE DETECTION LIMIT (ppb)
STANDARD DEVIATION (ppb)
MINIMUM DETECTION LIMIT
MAXIMUM DETECTION LIMIT
PHASE 4
AVERAGE DETECTION LIMIT (ppb)
STANDARD DEVIATION (ppb)
MINIMUM DETECTION LIMIT
MAXIMUM DETECTION LIMIT
SUMMARY
AVERAGE DETECTION LIMIT (ppb)
STANDARD DEVIATION (ppb)
MINIMUM DETECTION LIMIT
MAXIMUM DETECTION LIMIT
Chlorobenzene Chlorotoluene
N = 129
1.0
0.5
0.5
3.7
N = 133
0.7
0.2
0.4
1.6
N = 148
0.9
0.2
0.5
1.8
N = 155
0.7
0.2
0.4
1.5
N = 565
0.8
0.3
0.4
1.5

1.2
0.5
0.6
4.2

0.8
0.3
0.4
2.3

1.2
0.4
0.6
2:!-i
' 1
u.- ,
i
0.3
O.4
2.1

1.0
0 * i
A
0.4
2.1
                            910
-------
 FIGURE  1.  Equipment  Set-up for Performing  TAGA Calibrations
               Using  Standard Gas  Cylinders.
OUTSIDE AMBIENT AIM
                                        T/B" TEFLON TUBING
                                      ;J (>r LENGTH)
                                          GLASS SPLITTER TEES
 10ml/mln
TAGA INLET
PROBE
                                                 I


                                                 |
                                                 ^~
                    TAGA
                                         HALPERN SAMPLE AIR FLOW
                                         MEASUREMENT TRANSDUCER

                                                 BYPASS AIR
                                                                CHARCOAL ULTER
                                                                FOR EXHAUST
                                      SAF FLOW
                               ADJUSTMENT VALVE
                                          EXHAUST
                                                 AIR SAMPLING
                                                 PUMP
                  CYLINDER OF
                  STANDARD GAS MIXTURE
                                          911
-------
    FIGURE 2.  Equipment Set-up for  Checking the Transport
                 Efficiency of the TAGA Sampling Lines.
                       200'TEFLON LINED, HEATED
                       TRANSFER LINE
1.SOO nri/iK
OUTSIDE AMBIENT AIR
                                                                        TAGA
                                                        SAMPLE AIR FLOW
                                                        MEASUREMENT TRANSDUCER
                                                     BYPASS AIR
                                                                  CHARCOAL TRAP
                                                                  FOR EXHAUST
                                                 AIR SAMPLING PUMP
                                          912
-------
(O

CO
                                 FIGURE  3.     %  Transport  Eff.  vs.  Temp
                                               Unheated Sampling Lines, Phases 2-4
Percent Recoveries
118 -
116 -
114 -
112 -
110 -
108 -
106 -
104 -
102 -
100 -
98 -
96 -
94 -
92 -
90 -
88 -

+
o



J.
o +
o ° o + l
+ ° 4 A +o
/S V ^ T



4-0 + + * ^ 0
+ + ° 0
o
I f I I - — I 	 — I 	 1 	 — — . 	
                             30
                                     Chlorobenzene
    50

Temperature in F
           O
                                                                          60
                                                                      Chloratoluene
70
-------
QUALITY ASSURANCE FOR PERSONAL EXPOSURE MONITORING -
AN UPDATE
D. J. Smith
Analytical and Chemical Sciences
Research Triangle Institute
P. 0. Box 12194
Research Triangle Park, North Carolina
    The Research Triangle Institute has conducted personal monitoring
studies under the TEAM Program (Total Exposure Assessment Methodology)
since 1980.  These studies are complex, involving population studies and
sample selection, sample collection and analysis, and statistical analys*s
of the data produced.  A Quality Assurance Project Plan was prepared for
each of the studies.  This required careful planning for all study
activities including development of protocols and standard operating
procedures, complete documentation and review procedures, quality control
procedures for analytical systems, and quality assurance monitoring.  At
the end of each study, results and procedures are reviewed, and
methodology modified based on findings.  Some of these changes are
reflected in successive Quality Assurance Project Plans and some
from TEAM study activities are presented.
                                   914
-------
    The Research Triangle Institute has conducted personal monitoring
studies under the TEAM program (Total Exposure Assessment Methodology)
since 1980.  These studies are complex, involving population studies and
sample selection, sample collection and analysis, and statistical analysis
of the data produced.

    The basic study involves the collection of environmental and personal
air, breath, and water samples.   During the development of the TEAM Study
program,  methodologies for the collection and analysis of these samples
were evaluated and optimized.   Other methodologies have also been field-
tested and their usefulness for personal exposure monitoring evaluated.

    A brief summary of TEAM studies carried out by RTI is shown in Table
!•  The pilot study for the TEAM study and the first sampling trips were
carried out in two metropolitan areas in Northern New Jersey in Autumn,
!981,  Summer, 1982, and Winter,  1983.  Concurrently, two sampling trips
Were carried out in areas with little or no industry to provide low
background, or threshhold data,  in Autumn, 1982 and Spring,  1982.  In
1984,  TEAM sampling was carried out in several locations in California, in
Winter, 1984 (Los Angeles County), Spring, 1984 (Los Angeles County and
Contra Costa County).  More recently, a TEAM study was carried out in the
Baltimore metropolitan area by RTI and PEI Associates. Also in 1987,  two
peturn trips were made to Southern California to re-sample some of the
      where monitoring had been carried out in 1984.  And, most recently,
  TEAM study was carried out in Northern New Jersey; return visits were
     to sample some of the homes where monitoring had been carried out in
1980 - 1983.

    Other special studies were also carried out during the 1980's
Deluding:

         HEAL Study (Human Exposure Assessment Location)
         Nursing Mothers Study
         Dry Cleaners Study
         Swimming Pool Study
         Indoor Air Study

    From  the first pilot study through the present studies,  the cornerstone
of the sampling has been collection of air and breath samples on Tenax.  In
      phases of the TEAM program,  collection of water samples and analysis
   purge  and trap GC was an important and large part of studies.   The
         on water sampling has diminished, so that most of the focus  of
Duality assurance efforts is directed toward Tenax sampling.   Quality
Insurance procedures are implemented before and during the sampling period.
°efore beginning any sampling effort a Quality Assurance Project Plan is
         and protocols and SOPs  revised or created for methodologies  to be
       During sampling field blanks and controls are exposed and duplicates
^°Hected,  some of which are analyzed by a reference,  or QA  laboratory.
 Uring analysis of the Tenax samples by capillary column GC/MS,  daily
 hecks on calibration and instrument performance are performed.
 erformance evaluation samples,  prepared on Tenax by Environmental
MoiUtoring Systems Laboratory  (EMSL) of the U.S.  EPA or by an EPA
 Ofttractor laboratory,  are analyzed.

    In 1985 a report of quality  assurance procedures and recommendations
    improvements in the program  was published - "Quality Assurance for
                                   915
-------
Personal Exposure Monitoring," by R. W. Handy, H. L. Crist, and T. W.
Stanley  (1),  This presentation is an update of that 1985 report, in
particular, a look at the steps taken to address some of the concerns of
that report.

    First a look at some of the changes in the program itself.  Table II
shows the numbers and types of samples  (Tenax and water) collected during
the first three sampling trips to Northern New Jersey.  Originally,
approximately 3 1/2 samples were collected per participant - 2 personal a*r
and 1 breath sample, and outdoor fixed-site air samples at selected
locations.  The rate of duplicate sample collection was quite high - 30
percent of all personal and fixed-site air samples were collected in
duplicate; 25 percent of breath and water samples were collected in
duplicate.  Field control and field blanks were scheduled for approximately
10 percent of field sample collections; laboratory blanks and controls *et
analyzed as well.   Although this level of quality assurance was appropriat
for developing methodologies, 40% of the analyses were quality control of
quality assurance samples.

    Table III shows the same data for a more recent sampling trip.  This
1987 example shows a much greater number of samples per participant - 2
personal air, 3 indoor fixed-site air, 2 outdoor fixed-site air, 3 breath
samples, and water samples from only a subset of homes.  Duplicate sarapl6
collections were scheduled for 20 percent of the field samples; field
blanks and field controls were scheduled for 7 percent of field sample
collections.

    Maintaining high quality data with the current level of duplicate
sample collection and blank and control sample collection has been made
possible by improvements in Tenax preparation.  The 1985 report (1)
identified blank contamination as a major factor affecting data quality-
Subsequently, the protocol for cleaning and preparing Tenax was revised;
system for tracking Tenax by batch was implemented; QC criteria had to b*
met before Tenax cartridges were committed to sampling.  The improvement
field blanks can be seen in Table IV comparing mean blank background fr°*
Northern New Jersey, 1981 and California, 1984.  Improvements in the    .
storage of Tenax in the field and at the laboratory have also contribute
to maintaining low levels of background contamination.
                                                                       ,ejd
    Another area of concern identified in the 1985 report is that the ** j,
sample collection exceeded the analytical capacity of the laboratory. w"
results in samples being stored months before analysis.  A number of
advancements in technology have helped to increase analytical           ,0n
productivity - use of more efficient GC columns, use of better quantita
procedures and software.  Double shifts on instruments also increased
productivity.  However, the rate of sample collection in the field has
increased at an even faster rate.   The result is an even greater burden
the analytical laboratory.  Results of analysis of control cartridges.
especially those stored for long periods, still indicate minimum, if
loss of analytes.   Enough data are now available so that a detailed
statistical analysis should be possible to demonstrate the effects of
storage or analyte recovery, if indeed there are any.

    The results of performance evaluation sample analysis was the subJe
of lengthy discussion in the 1985 report.  Several instances of positiv
                                   916
-------
bias, and significant background levels of several analytes were cited, as
was variable precision as measured by relative standard deviation.

    Performance evaluation samples (Tenax) were assigned authentic sample
numbers and analyzed blind.  None of the data was corrected for background
or recovery and no data was rejected.  As background contamination of Tenax
cartridges was reduced, data quality improved; that is, precision as
measured by relative standard deviation improved and there was reduction of
bias for certain compounds.

    The performance evaluation sample program has changed since the studies
covered by the 1985 report.  EPA no longer analyzes cartridges along with
the contractor and the independent laboratory, and fortification levels are
Somewhat lower.  Table V shows some typical results for performance
evaluation samples.  The mean % bias and % relative standard deviation are
shown for each analyte.  For comparison, the * recovery for personal air
field controls for the same study is shown also.   In general bias and
recovery are similar for most analytes.  Also, the % RSD of the bias data
*s similar to median % coefficients of variation for duplicate personal air
samples for the study.  The exceptions are trichloroethylene (higher bias
&nd % RSD), and 1,1,2,2-tetrachloroethane and bromoform which were not
found in measurable amounts in significant numbers of samples.

    This program of providing Independently fortified Tenax cartridges to
    analysis laboratory is a particularly useful  one.  It would be even
     helpful to have analysis by EPA laboratories as a reference, or to
have are certified reference materials available.

    The TEAM program is presently alive and well.  New exposure monitoring
Programs and new methodologies are taking shape.   If there are lessons to
"e learned from TEAM monitoring, it is that ample quality assurance is
Ct%itical in early stages of development.  Without the information gained
*rom blanks, controls, and duplicates,  evaluating a developing methodology
    be most difficult, and comparison to other data impossible.
Inferences
l'   R.  W.  Handy,  H.  L.  Crist,  T.  W.  Stanley,  Special Technical Publication
    867,  ASTM,  Philadelphia,  PA (1985).

2l   Lance Wallace,  "The Total  Exposure Assessment Methodology (TEAM) Study:
    Summary and Analysis:   Volume 1," USEPA,  June 1987.
                                   917
-------
               TABLE  I.   TEAM  STUDIES  CARRIED  OUT  BY  RTI
       Autumn,  1981
       Summer,  1982
       Winter,  1983

       Autumn,  1982
       Spring,  1982

       Winter,  1984
       Spring,  1984
       Spring,  1984

       Spring,  1987

       Winter,  1987
       Summer,  1987

       Autumn,  1987
Northern New Jersey
Northern New Jersey
Northern New Jersey

North Dakota
Greensboro, NC

Los Angeles County
Los Angeles County
Contra Costa County

Baltimore, HD

Los Angeles County
Los Angeles County

Northern New Jersey
          TABLE II.  SAMPLES COLLECTED DURING TEAM STUDIES IN
                NORTHERN NEW JERSEY DURING 1981-19823
                                                         Samples
TEAM Study
Autumn, 1981



Summer, 1982



Winter, 1982



Participants
355 Field samples
Duplicates
Blanks
Controls
157 Field samples
Duplicates
Blanks
Controls
49 Field samples
Duplicates
Blanks
Controls
Tenax
1210
322
151
155
586
166
47
76
162
41
30
44
Water
717
188
57
59
306
70
17
17
96
25
15
1
aData from Reference 2.
                                   918
-------
         TABLE III.  SAMPLES COLLECTED DURING A TEAM STUDY IN
                           CALIFORNIA. 1987
TEAM Study
Participants
                                                         Samples
                Tenax
          Water
Winter, 1987
    51
Summer, 1987
    43
Field

Duplicates

Blanks

Controls

Field

Duplicates

Blanks

Controls
510

106

 35

 35

430

 60

 28

 28
18

 4

 3

 3

14

 4

 4

 4
                                919
-------
      TABLE IV.   COMPARISON OF TENAX FIELD BLANKS,  1981  AND 1984a
Target Compound
Chloroform
1 , 2-Dlchloroethane
1,1, 1-Trichloroethane
Benzene
Carbon tetrachlorlde
Trichloroethylene
jj-Dioxane
1 , 2-Dibromoethane
n.-0ctane
Tetrachloroethylene
Chlorobenzene
Ethylbenzene
Bromof orra
j>-Xylene
Styrene
(>-Xylene
1,1,2, 2-Tetrachloroethane
a-Pinene
£-Di chlorobenzene
j»-Decane
cj-Dichlorobenzene
jn-Undecane
ri-Dodecane
Field Blank
All Tenax
Autumn, 1981
(N=76)
22
1
33
97
2
3
NAC
NA
NA
11
1
12
ND
22
2
8
NA
NA
3
NA
1
NA
NA
Background, ng
Personal Air
Winter. 1984
(N=33)
2
NDb
6
17
ND
ND
ND
ND
ND
ND
ND
ND
NA
2
2
2
ND
ND
2
ND
3
ND
3
aData from Reference 2.
bND = Not detected.
CNA = Not an analyte for this study.
                                   920
-------
 TABLE V.  PERFORMANCE EVALUATION SAMPLE RESULTS, VOLATILE ORGANICS
                       WINTER, 1984 LOS ANGELES
Target Compound
Chloroform
1,1,2, 2-Tetrachloroethane
1,1, 1-Trichloroethane
Carbon tetrachlorlde
Trichloroethylene
Tetrachloroethylene
Bromoform
Chlorobenzene
Benzene
Ethylbenzene
fi-Xylene
Na
6
11
6
6
11
10
11
10
11
10
11
Mean
Bias, %
12.6
-14.6
17.7
9.9
43.7
12.2
-22.7
-16.7
31.2
-25.0
-13.5
% RSDb of
Bias Data
14.8
38.2
18.1
16.0
38.5
23.2
46.2
28.0
27.8
18.8
17.6
Field Control
Recovery
Mean * RECC
110
110
125
95
110
105
NAd
95
115
100
105
 N = Number of analyses.
b*RSD = * Relative Standard Deviation.
°*REC = * Recovery.
 NA = Not analyzed.
                                  921
-------
(XNSIDERATIONS IN THE DESIGSF OF AIR TOXICS M3SETORING
PROGRAMS AT SUPERFUND SITES
Richard Grume, Kent Kitchingman
Jeff Rosenbloom, Michael Stenburg, Arnold Den
U.S. Environmental Protection Agency, Region IX
215 Fremont Street
San Francisco, California  94105
    Ambient air monitoring activities at hazardous waste contamination
11 Superfund") sites nave recently increased as a result of a greater
within the Superfund program of the iitportance of inhalation as a signi
human exposure route.  As air monitoring activities have increased,
with the quality and representativeness of data generated have also become
more significant.  These problems, which relate more to sampling program
design than to execution, stem from three factors:  (1) inexperience of
personnel in the monitoring of air toxics, (2) the necessity of using non-
routine air monitoring methods, and (3) unclear intended data use.  Anoths
complicating factor is the ever-present need to obtain data as quickly as
possible.  We believe that the quality of air data collected at Superfund
sites can be improved through better "upfront" planning.  In particular,
there is a need for improved clarity in defining the intended use of data
prior to the onset of monitoring.  Additionally, more attention should be
directed towards the selection of appropriate sampling methods and the
determination of detection limits.  The use of pre-test screening models
also be useful in determining the potential significance of various
and exposure routes.  The purpose of this paper is to discuss the importa  (
of upfront planning in the design and execution of air monitoring programs
Superfund sites, and to present an air toxics monitoring checklist for
the initial planning phase of a project.  Special emphasis is placed on
assessment," since the ultimate use of air monitoring data is often to
risks to public health as a result of inhalation exposure.
                                    922
-------
introduction

     In assessing the health and environmental impacts from Superfund sites,
the U.S.  Environmental Protection Agency (EPA) must consider all potentially
Contaminated media,  including soil,  surface water,  ground water, and the
ambient air.   Air monitoring at Superfund sites is  relatively new  and
Untested compared to soil and water  measurements.   As  the frequency of air
"onitaring has increased in recent years, data quality problems have became
afparent.  Many of these problems have resulted from:   (l)  the inexperience
°f personnel in the  monitoring of airborne toxic compounds, (2) the use of
^n-routine methods  capable of measuring very low concentrations of airborne
toxic compounds, and (3) unclear intended data use.  Guidance in addressing
these problems during the design phase of an  air sampling program  is
Resented below.

Background

     The authority for the Federal Government to respond to releases,  or
threatened releases, of hazardous substances  that may  endanger human health
^r the environment is provided by the Comprehensive Environmental  Response,
Pptrpensation, and Liability Act (CERCLA), also known as the Superfund Law.
^is Law was enacted in December 1980 and amended in November 1986 by the
?^>erfund Amendments and Reauthorization Act  (SARA).   By Executive Order, EPA
ls given primary responsibility for  implementing the Superfund Law to clean
J$> hazardous waste sites or to respond to spills of hazardous substances.
r^ regulatory guidelines and procedures for  the Superfund  Law are described
j^ the National Contingency Plan.  EPA may take direct action to clean up a
^zardous waste site or, using the enforcement authority of the Superfund
     require those responsible for the release to implement removal or
         actions.
    An investigation at a Superfund hazardous waste site, known as a
      al Investigation/Feasibility Study (RI/FS), involves all environmental
      that may be contaminated through the release, or threatened release, of
  hazardous  substance.   These media include ground water, surface water, air,
    land.  A component of the RI/FS requires an evaluation of "risks" posed
   a population  as well as to the  environment.   (For example, the "maximum
      dual risk" provides an upper-bound estimate of the risk of cancer for a
^~kg  adult  exposed to the niaximum-occurring concentration of a toxic
r*npound for 70  years.   A maximum  individual risk of "10  " indicates that
-T^6 individual in a population of  1,000,000 may develop cancer during his
• lfetime.)   As those  involved in the Superfund program gain experience, it
    become apparent that the air pathway often represents a significant
       of risk.

    The purpose of air monitoring at  Superfund sites during the RI/FS
        is usually to characterize health risks associated with exposure to
      air contaminants,  both on-site and at any nearby residences, schools,
       and businesses.   Significant air  exposure risks may be associated with
    inhalation of:   (1)  volatilized toxics from contaminated soil, (2) wind-
      contaminated dust, and (3) toxic soil gases (e.g., contaminated
          Once  monitoring is complete,  air risks are considered along with
      risks  at the site (e.g., contaminated ground water), and an appropriate
         action  is selected (e.g.,  excavation of the site or the extraction
    treatment of contaminated ground water).  Some environmental scientists
      HI the value of risk assessment  as an absolute measure of the risk to
       health or as a factor in making cross-media risk comparisons; however,
   is  generally  agreed that risk assessment is a valuable tool in comparing

                                    923
-------
potential remedial actions, in establishing clean-up priorities, and in
allocating clean-up resources.

The Need for Upfront Planning

     The goal of Superfund air measurement activities is to obtain data
are representative and of high quality.  Although progress has been made
recent years , difficulties are still encountered in obtaining valid and
useful air toxics data.  These difficulties generally result from:

     o  Inexperience of personnel in the use of air toxics measurement
        techniques.

     o  The necessity of using non- routine air monitoring methods.

     o  Unclear intended data use.

A complicating factor is the ever-present necessity to obtain data as quic3a
as possible.  Fortunately, the above problems can be overcome through
improved upfront planning.  Several examples of how improved planning can
make the difference between success or failure during air monitoring at
Superfund sites are discussed below.

                              Intended Data Use

     We believe that one of the most critical elements in the upfront
planning of an air toxics sampling strategy is to develop a clear stated
of the intended use of the data by the Superfund site manager.  Some
decisions to be made in determining the intended use of data include:

     o  Whether to assess carcinogenic, systemic, or ecological effects.

     o  What populations to be concerned with {e.g. , the general
        children, or workers).
                                                                      -6 1
     o  What level of carcinogenic risk to be concerned with (e.g. ,10  '

     o  What type of risks to be characterized (e.g. , individual risk,
        population risk, or both) .

     o  Whether the measurement of air concentrations on-site, at the
        line, or within the surrounding community would be most
     o  What time period and geographical area should the measurement
        represent.

     o  What inputs are needed for dispersion modeling or emission rate
        calculations.
                                                                         f
The data usage decisions noted above all contribute to the determination
number of sampling program design parameters, including:

     o  Sample collection period.

     p  Duration of entire sampling program.

     o  Number of samples to be collected.

     o  Method detection limit.     __ .
                                    924
-------
    o  Compounds to be analyzed for.

    o  Location of samplers.

    o  Desired accuracy and precision of measurements.

    o  Use of average vs. worst-case  data.

    o  Acceptable meteorological conditions during sampling.
        all of the design parameters  listed above are dependent to some
      on the intended use of data by  the site manager, this relationship is
     ignored in the planning phase of air sampling projects.  For example,
   decision of whether  to be concerned with carcinogenic or systemic health
       can affect a number of  sampling parameters, as illustrated below.

                              Health effect to be assessed
                               Care inoqenic      Systemic

       Averaging period:           Annual    15-min, 8-hr, 2-wk
       Detection limit:           ppt-ppb           ppm
       Compounds to sample:     Carcinogens    Non-carcinogens
       Sampler locations:       Surrounding    Qn-site, near
                                 community         workers

       relationships also exist between other intended data uses and the
         sampling program design  parameters.  Thus, it is crucial that the
        use of monitoring data be well thought-out prior to selecting
        methods and designing  the test program.

   A related problem involves the interpretation and use of data by the
       especially when  the data are collected under non-representative
^itions.   For example,  carcinogenic risk calculations, which are usually
(w^d  on a 70-year exposure period, require as input annual average
J^entrations (including  annual averages estimated from shorter sampling
          However, in several  cases where data use was not considered
         I, carcinogenic risks  have been calculated using non-representative,
          data (i.e., data collected  over periods of hours, days, or weeks).
^s has resulted in a site manager being put in the difficult position of
          to explain to the public results showing high cancer risks,
   these same risks calculated from  data collected over a representative
           may not have been alarming.  In these cases, the public should
   have been promised a risk assessment based on non-representative exposure
  "l»  and an appropriate use for the short-term exposure data should have been
          before going  into the field.

                       Selecting  Appropriate Methods

   Once the intended use of the  data has been determined, the design
          discussed above can  be  defined, and appropriate sample collection
   analytical methods can be selected.  Tne selection of these methods
          a critical part of any  air  toxics measurement plan and must be
     on the specific data needs and conditions imposed by the site.  For
        if a solid sorbent is  used, its adsorption and desorption
            for all the compounds of  interest at the site must be known,
^ if the sorbent has  been used  successfully in other situations.
         this was not done prior  to the sampling of a number of organic air
         its.   Instead,  spiked  tubes (i.e., sorbent tubes exposed to known

                                   925
-------
quantities of contaminants) were sent to the laboratory simultaneously with
the field samples, such that adsorption/desorption efficiencies would not be
determined until the field samples were analyzed.  The results from the
spiked tubes indicated that the sorbent was totally ineffective for the
measurement of several compounds which were important to the study.  This
resulted in an unnecessary data gap in the project.

     One factor that is frequently overlooked during the initial
determination of sampling methods and procedures is the relative toxicity of
the individual contaminants present in the sampled air.  For example,
sometimes sampling parameters  (i.e. , flow rate and duration of sampling) are
based on the expected predominance of one compound (e.g. , based on expected
mass concentration) , overlooking other compounds that are present at much
lower concentrations but which still exhibit significant risk due to their
high carcinogenic potency.  This is illustrated in the example below where,
although 1 , 1-dichloroethylene is present at just one-hundrellC
Health Evaluation Manual gives additional advice on the selection of target
compounds for monitoring purposes.

                               Detection Limits

     Perhaps the most obvious example of the need for upfront planning is in
the determination of required detection limits.  The detection limits for &
project are defined by the intended use of the data (e.g. , estimating the
excess cancer risk resulting from air releases from a waste site).  The
lowest risk of concern to the decision-maker must be identified upfront _
before detection limits can be calculated, and before appropriate sampling
and analytical methods for the project can be determined.   Unfortunately/
there is a tendency to select sampling and analytical methods before the
use of the data has been defined.  When this occurs ,  the resulting data i
be unusable or of limited use because the levels of detection were too
to meet the program objectives.  Alternatively, program resources may tie
wasted as a result of using a method with detection limits lower than
necessary.

     As an example of this, a recent sampling plan called for the use of a
portable photoionization instrument to assess ambient concentrations
trichloroethylene.  Since the detection limit of -this instrument
to a maximum individual risk in the range of 10   for trichloroethylene,
instrument was clearly inadequate for the assessment of risks in the ran9e
interest (in this case 10  ).  Similar problems have occurred where the
sampling method was correctly chosen, but the planned sampling duration a£°
volume were not, resulting in the collection of insufficient sample mass to
obtain the desired detection limits.

     While detection limit problems can usually be overcome through car .J^j.
planning, sampling at the desired detection limit may still not be practic

                                    926
-------
for some compounds having relatively high carcinogenic potencies.   For these
compounds, several approaches can be taken to minimize ambient dilution, and
thus, increase the likelihood of detection.  One approach for gases being
released from solid or liquid surfaces is to make ground-level measurements,
possibly with the aid of a flux chamber.   Another approach would be to
sample during periods of high atmospheric stability, when contaminant
concentrations in ambient air are greatest.  However, either of the above
approaches requires caution in extrapolating results to representative
exposures (e.g., breathing zone concentrations and typical atmospheric
conditions).

Risk Screening Approaches

     At the onset of a Superfund site investigation, the routes of
significant exposure may not be known (e.g., inhalation of contaminated
vapors or dusts, ingest ion of contaminated drinking water or soils, or dermal
exposure to contaminated soils or water).  Since monitoring costs and time
constraints limit the assessment of all potential routes of exposure,
decisions must be made regarding which routes are most likely to represent
significant health risks.  In some instances in the past, these decisions
have been made based on limited information, leading to errors in the
selection of potentially significant exposure routes for investigation.  A
preferred approach would have been to make use of several simple risk
screening models, as described below.

     An example of a risk screening approach is EPA's model for assessing
emissions from PCB-contaminated soil.   This model predicts the maximum
individual risk for exposure to volatilized KB compounds as a function of
soil contamination level.  The model suggests that at low contamination
levels, the risks from the inhalation of vaporized PCBs may be too low to
Warrant the expense of air monitoring.  Alternatively, at higher soil
contamination levels, air monitoring may be worthwhile.  (See Figure 1.)  The
model was recently used in the Region IX Office to predict the potential
health impact associated with vapors from soil contaminated with an average
of_200 ppm_qf PCBs (Aroclor-1242).  The predicted risk was in the range of
10 4 to 10  .  Later, when preliminary air monitoring was performed, the
actual risk was found to be in the range of 10   to 10  .  Thus, the model
Predicted actual risks within an order of magnitude, erring on the side of
health protection.  This magnitude of error is considered good for screening
anaylses of this type.  Similar models can also be used to assess the
Potential impact of other volatilized soil contaminants.

     Another useful risk screening approach can help to determine whether
resources should be allocated to the monitoring of wind-blown dust.  The
levels of dust contamination required to cause an on-site maximum individual
*lsk of, say, 10   can be approximated by making the worst-case assumptions
that dust concentrations are present for a lifetime, 24 hours/day, at the
level of the former National Ambient Air Quality Standard for total
Particulates of 75 ug/m .  Using this approach, and assuming that soil (and
therefore dust) contamination levels are present at realistic values, it can
tie demonstrated that the maximum individual risk associated with the
inhalation of wind-blown dust for most common Superfund soil contaminants is
s*tremely low (i.e., well below 10   for most compounds).  Thus, the
'Measurement of wind-blown dust concentrations is usually not a worthwhile
Denture.  However, dust measurements may be justified for several common soil
contaminants having high toxicities.  For example, using the above worst^case
^sumptions, the soil contaminant concentrations required to cause a 10
Airborne risk from the inhalation of wind-blown dust are shown below for a
       of compounds having high toxicities.

                                    927
-------
                                     Soil concentration

               Arsenic                       3 ppm
               Beryllium                     6
               Chromium VI                   1
               Dioxins/dibenzofurans         0.0004
               Polychlorinated biphenyls    10

 (Similar calculations can also be made for other compounds and risk levels of
 interest.)  In situations where highly-toxic conpounds are present in the
 soil at levels similar to those listed above, and where the contamination
 appears to be widespread, the monitoring of wind-blown dust may be
 appropriate.

     The above approach was recently used to estimate risks at a site having
 soil contamination ranging from 40 to 7,000 ppm of arsenic.  MDnitoring_4
 results showed that actual risks, which were in the range of 10   to 10  ,
 agreed reasonably well with the predicted risk range of 10   to 10  .

 Quality Assurance

     Quality assurance consists of upfront planning, ijiplementation, and
 post-analytical activities.  Although this paper has emphasized upfront
 planning as a means of obtaining data that meet program objectives,
 should be paid to all aspects of quality assurance.   For example, there is a
 special need at Superfund sites for the site manager to be aware of the
quality of his data.  This can be accomplished through data validation
 techniques.

Air Toxics Monitoring Checklist

     In an effort to improve future Superfund air monitoring activities,
have prepared a checklist of items to be considered in preparing sampling
quality assurance plans.  The checklist, which is presented in Appendix I,
 emphasizes those elements that are often neglected in the planning stage of
 air monitoring, but nevertheless are critical to the success of air toxics
measurement activities at Superfund sites.  It is hoped that the checklist
will be useful to Superfund site managers and contractors while planning
toxics measurements.

Conclusions

     The importance of upfront planning during the development of Superfund
air measurement plans has been discussed.  Risk assessment may play an
 important role in these planning activities, especially where the objecti^6
of the sampling effort is to characterize risks to human health as a result
of inhalation exposure.  In these cases, it is recommended that the service^
of experts in the areas of quality assurance, air toxics, and toxicology I3®
used to help define project objectives, to provide guidance in the collect*01
of appropriate data, and to assist in the assessment of risk.
                                    928
-------
References
3.
"Superfund Public Health Evaluation Manual," U.S. Environmental
 Protection Agency, Washington, B.C., EPA-540/1-86-060,  1986.

M.R. Kienbusch, W.D. Balfour, S.J. Williamson,  "The Development of  an
Operations Protocol for Emission Isolation Flux Charriber  Measurements on
Soil Surfaces," Proceedings of the 79th Annual  Meeting of the Air
Pollution Control Association, Minneapolis, Minnesota, 1986.

"Development of Advisory Levels for Polychlorinated Biphenyls  (PCBs)
Cleanup," U.S. Environmental Protection Agency, Washington, D.C., CHEA-E-
187, 1986.
     -2
         .00
                            -6.00
                                              -5.00
                                                                       -4.00
                               LOG (Inhalation Risk)
                    D   Kd = 1,000              +    Kd - 40
figure 1.   Cn-site lifetime PCB3irihalatian risk as a function of soil PCB
•Qnqentration for Aroclor-1242.   3Vo soil-water partition coefficients, K,
     water / g soil) , are illustrated.  Values are corrected to reflect a
       potency factor or 7.7 (mg/kg-day)  .  (For other Aroclor compounds,
         routes , and exposure distances , see Reference 3 . )

                                     929
-------
                                                              APPENDIX 1
                    AIR TOXICS MONITORING CHECKLIST
Site:  	
EPA site manager:  	
Sample team manager and affiliation:
Scheduled field sampling date:  	
Datauser(s):  	
                   (Check All Applicable Boxes)
Intended use of monitoring data:

     [  ]  Human exposure:                   [  ]
          [ ]  Community exposure
          [ ]  Worker exposure
          [ ]  Children or elderly

     [  ]  Carcinogenic health effects:
          t ]  Max. individual risk
          [ ]  Max. exposed individual
          [ ]  Population incidence
          [ ]  Specify risk level of
               interest:  	

     [  ]  Systemic health effects          [  ]

     [  ]  Ecological effects

     [  ]  Determination of action level:    [  ]
          [ ]  Minimum risk
          [ ]  Ambient concentration
          [ ]  Other:  	
                                                Other uses:
                                                t  ]  Fence-line cone.
                                                t  ]  Odor assessment
                                                [  ]  Disp. modeling
                                                [  ]  Em.  rate calc.
                                                [  ]  Upwind/downwind
                                                     comparison
                                                [  ]  RI/FS remedy
                                                     selection
                                                [  1  Other:   	
                                                Type of analysis:
                                                [ ]  Screening
                                                [ ]  Detailed

                                                Type of estimate:
                                                [ ]  Ave. or typical
                                                [ ]  Worst-<;ase
Based on intended use of data, determine the following parameters;

     [ ]  Compounds for measurement
     [ ]  Sampling period
     [ ]  Method detection limit
     [ ]  Duration of sampling program
     [ ]  Number of samples to be collected
     [ ]  Sampling locations (including background monitoring)
     [ ]  Data quality objectives (e.g., accuracy and precision)
     [ ]  Sampling and analytical methods
                                   930
-------
Representativeness of sampling strategy;

     [ ]  Account for seasonal and diurnal effects

     [ ]  Define acceptable range of process and meteorological parameters
          during testing

     [ ]  Record process and meteorological parameters, and soil
          temperatures (where appropriate), during testing

     [ ]  Can collected data reasonably be extrapolated to the required
          averaging period?

     [ ]  Assess background concentrations, where appropriate

     [ ]  Do the number and location of samplers take into account:
          [ ]  Local variations in soil concentrations
          [ ]  Local meteorology
          [ ]  Intended use of data (e.g., conmunity exposure)

     [ ]  Will ground-level or breathing-level data be required?


In assessing risks f will the following factors be considered:

     [ ]  Changes in contaminant levels with time
     [ ]  Expected population growth
     [ ]  Anticipated exposure period (if < 70 years)
Will monitoring and modeling techniques be comparable with those
Used to assess water and soil exposure risks, with respect to the
use of;

       ]  Average vs. maximum concentration values
       ]  On-site vs. off-site concentrations
       ]  Dilution factors
       ]  Projected vs. actual concentrations
       ]  70-year vs. shorter exposure duration
     [ ]  24-hour vs. shorter exposure frequency
Mscellaneous concerns;

     [ ]  Perform preliminary exposure route risk screening,  where
          possible

     [ ]  site safety requirements should be consistent with  potential
          human exposure
                                   931
-------
STATISTICAL ANALYSIS OF GC/MS PERFORMANCE
AUDIT DATA
Raymond C. Rhodes
Howard L. Crist
Quality Assurance Division
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina
     Several sets of GC/MS data  from performance audits at different labo"
ratories were analyzed.  The  organic compound data that were analyzed wefe
displayed in  graphical plots.   Significant  patterns emerged  that  suggest
different types  or  sources  of experimental  measurement error.  Investiga'
tion of the possible causes of the errors  could result  in further improve"
ment of data quality for  performance  audits and analyses of routine samples*
                                   932
-------
Introduction

     An important  activity  of any quality  assurance  program is  conducting
performance audits.   Performance  audits  are  used to  obtain  quantitative
data on the  accuracy of the measurement  system.   A common procedure  is  to
prepare audit  samples  in  the  most  accurate  and  precise  manner  possible
using the best standards available and including  them in  the  measurement  or
analytical process  as  blind  samples  or unknowns.  The actual levels of the
concentrations to  be measured should  be  near the  levels of routine  field
samples.  The  analyst  should process the performance  audit samples in the
same way as routine  samples and, ideally, the analyst should  not  know  which
samples are the performance audit samples.

     Increasingly, studies are  being conducted  to sample  for  and analyze
volatile organic  compounds (VOC's)  in  the  air.   A widely used approach  of
monitoring for VOC's is by  collection on  a  solid sorbent  such as   Tenax
followed by thermal  desorption  and combined  gas  chromatography/mass  spec-
trometry (GC/MS) analysis.

     Because of  possible  interferences   among  the  organics  of   interest,
Various combinations  of high  and  low  concentrations  should  be  used   in
preparing the  audit  samples.  However,  to  prepare many  such  combinations
^ould be a considerable effort.  Usually, to minimize this effort,  a master
gaseous mixture of the organics is first prepared, and then various amounts
of this mixture are spiked onto the Tenax cartridges.

     In this paper,  the  current standard practices  for  analyzing perform-
ance audit data are briefly reviewed, and recommendations are made for pre-
senting the data  in  graphical form and for further statistical evaluation
°f the  data.   A  recommended  procedure  for  evaluating and  analyzing  audit
results is included.

Evaluation of Results of Performance  Audits

     Data from a  recent audit  are  used as  an  example.   The spiking levels
of samples of nine organics were:

                                        Nanograms
                                  Low    Medium    High

          1,2-dichloroethane       114      170     227
          1,1,1-trichloroethane   122      183     244
          benzene                  80      119     159
          carbon tetrachloride    145      217     289
          trichloroethylene        133      199     267
          tetrachloroethylene     147      221     294
          chlorobenzene           101      151     201
          ethylbenzene             79      118     157
          o-xylene                 80      119     159

     separate  samples at each level were prepared  for  analysis by  one  labo-
        The  detection limits for the above  chemicals  vary from 1  to 4  nan-
°grams.

     Standard  practice  for  performance audits  has  been  to  evaluate  the
        by  computing the percent difference between the  observed analysis
value and  the  known spiked value.


                                    933
-------
            Percent Difference = Observed - Known  (100)
                                      Known

Percentage differences are  usually computed  for  chemical analyses  because
the errors  are  generally proportional  to  level.   The  results  were summa-
rized by  calculating  the  average  and  standard  deviation  of  the  percent
differences.
                                                   Standard
                                  Average, %     Deviation,^

          1,2-dichloroethane        -20.8           20.9
          1,1,1-trichloroethane     -30.7           23.6
          benzene                     3.2           29.1
          carbon tetrachloride      -33.5           18.0
          trichloroethylene         -25.4           23.4
          tetrachloroethylene        -9. A           28.6
          chlorobenzene               7.2           34.3
          ethylbenzene               14.7           37.5
          o-xylene                    9.9           37.8

The average percent difference is a measure of bias; the standard  deviation
is a measure of precision.  This is as far as many evaluations go.   The  re-
sults are submitted to the project supervisors and the laboratory  involved.

             Statistical Evaluation of the Percent Differences

     A number of statistical tests can be performed on the results and some
interpretation of the results can be made.  Tests can be performed  to deter*
mine if the  biases  are  statistically  significant and  to  determine if  the
precisions are essentially the same.  Also, confidence limits can  be placed
on the biases and the precisions and on future individually reported values
or averages.

                      Graphic Presentation of Results

     Much can  be learned  from  graphic presentations  of  data.   In  most
cases, graphic presentations  should  be evaluated  prior to  any statistical
analyses for several  reasons.  First,  outliers  are more  readily evident-
Questionable values that do not seem to fit in with the predominant  porti°n
of the  data  should  be  double checked  and  validated,   if  possible,  bef°re
proceeding.  Statistical tests  can be used to provide  a  more quantitative
evaluation of the potential outliers.  Decisions must  then  be made
or not to include the outlier values in the analysis.   (It could be
to perform the  statistical  analyses  both  with and without  the outliers-
In some cases  it may  be desirable to  replace the outliers  with expects
values in performing the statistical analysis.
     Second, patterns of  the  data are much  more  readily discernable
ally from  graphic  presentations  than  from  the  numerical  values.
nition of such patterns  is  helpful in determining  the  need to make
formations of the data  and  in making a  final  decision  on the assumed
tistical model to be used in the analysis.

     For performance audit data, a useful graphic presentation is  a plot °
the observed  values  versus  the known  values.   Examples  of  such plots &*
shown in Figures 1  and  2.  Note  that  the  sample code  letters  are used
the plotting points.  Without using the sample code information, regress *
lines could be computed for each  chemical; however, the  significance of c

                                   934
-------
sample code letters will presently be shown.  The line of perfect agreement
is shown as  the dashed  line  with a  slope  of  one  and  intercept  of  zero.
Note in Figure  1  the intercept of  the  regression line  is  near zero (only
slightly positive) and  the  slope  somewhat less than one.   In  Figure  2, it
appears that  the  intercept  is considerably  more than  zero  and  that  the
slope is considerably less than one.  The scatter of the points in Figure 2
at given known  levels  is  about  the  same  regardless  of  observed  level.
Figure 1 indicates a slightly smaller amount of scatter at the  lower level.
The assumption  implicit  in  the calculation of  percent  differences is that
the intercept of  the  regression  line  should  be  near  zero  and  that  the
amount of  scatter  of the  points  about  the  line should  increase  linearly
with increasing level of  the  spiked  compound.   Figure 2 indicates a depar-
ture from  these  conditions,  implying that the  assumptions  for calculating
percent differences may not be valid.

     Another observation  that can be  made  from  these  two  figures  is  the
similarity of the pattern of the sample  code  letters.  The  similarity of
the pattern was even more striking  when other  chemicals of this study were
considered.  This  is  strongly  evident  of  significant  saraple-to-saraple
differences, even at the same known level.

     The above  observations  indicate  that  results  of the performance audit
should be analyzed and summarized in a  way different  from simply calculat-
ing percent differences.

Recommended Procedure for Evaluating and Analyzing Audit Results

     The following procedure for evaluating and analyzing performance audit
     is recommended.

     1.  Plot the observed versus known values for each chemical in a man-
         ner similar to that  shown in Figures  1 and  2  with the sample code
         letters.

         If any  outlier  values seem  to be  present at a  given level, test
         the outlier value by using  Grubb's test.  If  it  is a significant
         outlier after investigation  and  validation,  delete the value from
         further analysis.

     2.  Perform a regression analysis for each chemical.

         Test the slope  for  significant  difference  from  1.0.   (The  slope
         should  obviously be  significantly different  from zero; otherwise,
         a new method of measurement should be  developed.)

         Test the intercept for significant difference from zero.  If it is
         not significantly different  from  zero,  recompute  the regression
         line, forcing the line through the origin.

         Compute the standard  error  for the  regression.   This  is a measure
         of the  scatter of the points about the line.

         Compute the deviations of each  point from  the regression line, not-
         ing the sample code for each.

     3,  If a pattern of the sample  code letters appears similar across the
         plots for the  chemicals, perform an  analysis  of  variance  on  the
         deviations to  determine  the  significance  of  differences  due to
         samples.
-------
          If  significant, compute  the variability due to samples.  If sample-
          to-sample  variability  is significant,  the cause of the variability
          should  be  investigated.   If  the sample-to-sample variability could
          be  eliminated  or  reduced, the  variability  of  the  measurement  sys-
          tem as  estimated  from  the   performance  audit  results  could  be
          appreciably  reduced.

Proposed  Design  for Future Audits

     It was  noted  earlier  that the low  samples were low for  all  chemicals
and  similarly  for  medium  and  high  samples.   If  any  interaction  effects
exist, the design will not  reveal them.  If any  interaction effects are sus"
pected to exist, other designs should be used, even though it  may  require
more effort  in the preparation  of the  samples.   If specific  interactions
are  suspect, particular combinations of  these chemicals should be built  into
the  design.  For example, if  interaction effects  are suspected only  between
chemicals 1  and  2  of the 9  chemicals, the following design could be used:
Sample
A
B
C
D
1
L
H
L
H
2
L
H
H
L
3
L
H
L
H
Chemical
456
L
H
L
H
L
H
L
H
L
H
L
H
7
L
H
L
H
8
L
H
L
H
9
L
H
L
H
     where H is high level
           L is low level

     As many replicates of the above four samples could be employed as costs
and time permit.  Three  complete  replicates would require only 12  samples*
the same as  used  in the performance  audit  described.   A  check for one of
two additional interactions could be  built  into  the design without  appreci'
ably increasing  the total  number  of samples.   However, if  all  possibl6
interactions were  included  in the  design,  the  number  of samples  required
would be prohibitively large.  Of  course,  there  may  be good chemical  tech*1
nical reasons why many of the interaction effects are  considered  impossible
and would therefore need not be included in the design.

Conclusions
     More extensive  evaluations  and  analyses   of   the  results  than
usually been done have revealed the presence of  outlier values, signifies!1*
deviations from the assumed model, and significant saraple-to-sample differ"
ences.

Recommendations

     The following  recommendations  are made for  conducting audits and f°f
analyzing results from performance audit  samples for GC/MS analysis of of
ganics:
1.
2.
         Include in  the  design the spiked  levels  necessary to investigate
         any strongly suspected interactions,
         When reporting  and  evaluating the results  of  the audits,
         the results in  various  graphic forms and  perform various stati6'
         tical analyses of the results, including:

                                   936
-------
         a.  Tests for outliers,
         b.  Tests for significant sample-to-sample differences,
         c.  Regression analysis.

     3.  Develop  an  interactive computer  program to plot the  results  and
         analyze the data as suggested herein.

Note:  The above recommendations apply to any performance audits in which a
number of chemicals are analyzed from the same sample.
                                   937
-------
   260
                     50
                                    100             ISO

                                      KNOWN VALUES
                                                                    200
ISO
                  Figure 1. Observed vs.  known values for 1.2 dichloroethane.
  250
                                           LAB A
  200
                                                                                 OC
o
!S
2
0)00 -
  so -
                                                             OB
                    SO
                                   100             160


                                     KNOWN VALUES
                                                                   200
                Figure 2. Observed vs. known values for 1,1.1-tricnloroethane.
                                                                                  250
                                        938
-------
QUALITY ASSURANCE FOR MEASUREMENTS BY EPA METHOD 5G
FOR WOOD HEATER CERTIFICATION TESTING
Glenn D, Rives and Michael W, Hartman
Radian Corporation
Research Triangle Park, North Carolina  27709

Thomas E. Ward
U. S. EPA/EMSL-RTP/QAD/Source Branch (MD-77A)
Research Triangle Park, North Carolina  27711


     The U. S. Environmental Protection Agency (EPA) has recently
promulgated New Source Performance Standards (NSPS) regulations for
particulate emissions from wood-burning heaters that require testing of
new wood heaters to demonstrate compliance with the emission limits.
The regulations permit testers to choose one of two emission sampling
approaches:  EPA Method 5G (dilution tunnel) or EPA Method 5H (stack
sampling).  For reasons of cost and relative ease of performance, the
sampling method predicted to be used most often is EPA Method 5G.  This
paper presents an evaluation of the Method 5G sampling and analytical
measurements and presents the quality assurance (QA) measures necessary
to determine quality of the data produced.

     Quality assurance during wood heater testing by improving consistency
and reproducibility of data can be a major component in accomplishing the
goal of the NSPS, that is reduction in concentration of ambient particu-
late matter.  This paper will focus on EPA Method 5G.

     The Method 5G specification for dilution tunnel flow rate (140 scfm)
corresponds to velocity head readings of approximately 0.037 in.  H_0 using
a standard pitot, and 0.044 in. H.O with an S-type pitot tube.  Such low
velocity head readings are on the measurement threshold of commonly used
measurement instruments.  Degradation of measurement accuracy and
precision occurs as velocity head readings approach these minimum
detectable limits.

     Total particulate catches measured by using Method 5G sampling trains
are commonly in the range of 5 to 50 rag.  The lower portion of this range
of gravimetric catches is approximately an order of magnitude smaller than
amounts typically collected by using other source sampling methods.  Such
small total catches require special precautions and care in performing
train cleanup, desiccation, and weighing procedures.

     Extremely small errors in the measurement of either the velocity head
or gravimetric catch result in significant errors in the final calculated
emission rates.  Quality assurance and quality control procedures are
recommended to improve the quality of each of these critical measurements.
                                   939
-------
Overview of EPA Method 5G

     Method 5G is applicable for sampling wood heater particulate
emissions from a dilution tunnel location and is used with EPA Method 28,
"Certification and Auditing of Wood Heaters," for determining compliance
with the New Source Performance Standards (NSPS).  Method 5G describes
specifications for constructing a dilution tunnel for capturing wood
heater exhaust and mixing with ambient dilution air.  The method also
describes specifications for three sampling train options, each with
unique configurations, specifications, and sample recovery procedures.

     Figure 1 contains a schematic of a dilution tunnel sampling system
constructed according to the guidelines in the method.  One of the main
reasons for using a dilution tunnel sampling location is that such an
approach allows a direct measurement of flow rate.  The low flow rates
typical of wood heater exhaust stacks (-5 cfm), make direct measurement of
flow rate using conventional pitot tubes and differential pressure gauges
impossible.  A second reason for a dilution tunnel sampling approach is
that it provides sufficient lengths of straight ducting to allow selection
of a sampling location that is subject to minimal flow disturbances.

     In addition to the dilution tunnel specifications, Method 5G
describes specifications for three sampling train options for use in
measuring particulate matter concentrations.  Method 5G measurements and
QA procedures discussed in the following sections are applicable to each
of the three 5G sampling options.

Identification of Method 5G Measurement Variables That Have the Greatest
Affect on Data Quality

     Of all the measurement variables involved in testing with Method 5G,
two have the most potential to have a significant affect on data quality.
These are the volumetric flow rate determination and gravimetric
determination.  Quality assurance/quality control guidelines for these two
measurements are necessary because errors directly affect the emission
rate calculation.  Both of these measurement variables are sensitive to
factors such as operator technique, equipment type,  and operator
knowledge.  A brief discussion of these two variables follows.

     Measurement of flow rate as prescribed by Method 5G, although much
improved over many stack measurement approaches, remains a very sensitive
measurement and subject to error.  In developing the dilution tunnel
sampling approach, considerable effort was given to designating a flow
rate that is large enough to measure.  The opposing difficulty is that as
flow is increased, particulate concentration in the tunnel decreases, and,
therefore, the mass of particulate collected in the sampling train
decreases.  Figure 2 shows the relationship between dilution tunnel flow
rate and particulate matter collected in a Method 5G sampling train when
emissions are sampled from a catalytic heater performing at the level of
the standard.  The flow rate designated by the method (140 + 14 cfm) is
the optimum condition for both a measurable A? value and a measurable
sample catch.  Note, however, even at this optimal condition, both
measurements are on the low end of the range that is commonly considered
measurable.  The designated flow rate corresponds to a ^P value of
                                   940
-------
approximately 0.037 for a standard pitot and 0.044 for a type S.  Total
particulate matter catch is dependent upon the length of the test run but
is commonly in the range of 25 mg and often, for very well controlled
heaters or for short burn durations, the gravimetric catch is less than
10 mg.  An added difficulty in measuring these small particulate catches
is that these values represent total train catch, that is, the sum of
three train components;  two filters and the probe assembly.  Accurate
measures of these particulate catches are further affected by the moisture
factor, as well as any volatility of the organic matter in the sample
train catch.

Recommended QA Guidelines

                   Dilution Tunnel Flow Rate Measurement

     1.   Use an S-type pitot tube rather than a standard pitot tube.   An
S-type pitot gives an approximately 20 percent higher pressure differential
reading on the nanometer.  Also, the S-type is less sensitive to alignment
errors than a standard pitot tube.

     2.   Use a microtector manometer.   Microtectors allow the measurement
of extremely small ^p values.  Over the ranges necessary for dilution
tunnel velocity measurements, these instruments show greater accuracy and
precision, and have much better resolution than conventional differential
pressure gauges.

     3.   Perform both pretest and post-test traverses using four points
Per traverse.   Only 4 traverse points are used in the 6-inch tunnel
diameter.  More than four points would cause the outermost points to be
too close to the tunnel wall creating a bias in the measurement.  Extreme
care should be taken when performing tunnel traverse measurements.  Proper
pitot alignment and the allowance of sufficient time for the manometer to
respond should be assured during measurements at each traverse point.   A
minimum of 30 s and preferably 1 min should be allowed at each point for
the differential pressure reading to stabilize.  A post-test traverse
should be performed to correct for changes in tunnel flow characteristics
that occur over the course of a test run.

     4.    Place the pitot tube at the centroid of the dilution tunnel for
£est run measurements.   The center location is optimal for pitot tube
placement because it is influenced the least by wall effects and generally
results in the largest differential pressure reading.  Use the average
tunnel velocity derived from the pretest velocity traverse to calculate a
center adjustment factor.  After the test run,  perform another velocity
traverse.  If the center factor determined from the post-test traverse
differs from the pretest center factor,  use the average in the flow rate
calculations.

     5.    Perform an independent check on tunnel flow rate and pitot tube
Calibrations by using a dimensioned orifice.    This dimensioned orifice
should be placed between the suction fan and the dilution tunnel and the
pressure drop across the orifice measured.   The pressure drop is directly
related to the velocity of the gases.  This check serves to verify both
the tunnel flow rate and the pitot tube calibration.
                                   941
-------
                         Gravimetric Measurements

     1.   Use a calibrated analytical balance that reads to five places.

     2.   Perform all weighings in a temperature and humidity controlled
area.   Moisture effects have the potential to create a significant bias
on the small quantities of particulate collected in each of the sample
train fractions.  For this reason, room conditions should be carefully
controlled and the length of time that sample components are out of the
desiccator during weighings should be limited.  Failure to dry the filters
and sample containers before weighings (tares) is a major contributor to
erroneous results.  This is of special concern for the 5G sampling option
that allows direct weighing of the probe in determining sample collection.

     3.   Use identical procedures for pretest and post-test weighings.
Always treat sample train components identically during pretest and
post-test weighings.  Preferably, the same experienced technician should
perform all weighings.  Closely monitor the room conditions, the length of
time that sample components are desiccated, and the length of time that
samples are out of the desiccator during weighing.  All of these
conditions should be the same for pretest and post-test weighings.

     4.   Practice meticulous cleanup procedures.   Recovery of sampling
train components should be performed with extreme care by an experienced
technician.  Special care should be taken while handling the sample
components to keep from introducing contaminants or from losing the
sample.  Special care should be taken while recovering filters to ensure
that all filter fibers that stick to the filter holder gasket are
recovered and included in the post-test weights.

     5.   Always use sample component blanks to verify weighings.    To
ensure that the measured sample catch weights are not biased, sample
component blanks should be used.  This also serves as a measure of the
effectiveness of the cleanup and recovery procedures.  For example, if
weights on the blank filter differ by more than 1 mg between the pretest
and post-test weighings, the recovery and weighing procedures should be
reexamined, particularly the recovery of filter fibers from the filter
holder gasket.  For sampling option 5G-3,  the probe as well as the filter
should'be checked by using blanks.  In general, testers should set up a
complete blank train and treat it the same as the sampling train.
Leak-check the blank train, and recover the blank sample by using the
procedures as for the sampling train.

Disclaimer

     This paper has been reviewed in accordance with the U. S.
Environmental Protection Agency's peer review and administrative review
policies and approved for presentation and publication.
                                  942
-------
                                    Exhaust

                                    A
         90 'El bow
                 6*- 12" Baffles
           >V^
         h-mln12*-H
         vTTXJsTTrif./Tj Usff*
          'sassi
           Stove
           Sgale
                       Sample Port ,
                        Location ^r~l
^
                   Sample Point Location
                     (center of stack)
  Damper
Elbow
s
.
Velocity Traverse
Ports
«*nn*lnn •-
!
i
c
minimum
I
?
I
I
v
->
£
/ J



?M
it
                    Figure 1.
                                       Blower
           Schematic of the
Dilution Tunnel Sampling System
                     943
-------
Relationship Between Dilution Tunnel
Flow Rate, Sampling Train Particulate
     Catch, and Velocity Head (Ap)
    120n
                       • mg Paniculate Catch

                    	Type-S Pitot Tube

                    • • • • Standard Pitot Tube
0.160
                                    0.140
                                    0.120
                                    0.100
                                    0.080 -.
                                    0.060 O
                                    0.040
                                    0.020
       0    50    100    150   200   250

         Dilution Tunnel Flow Rate, DSCFM

                   Figure 2.
                    944
-------
 AN ALTERNATIVE STANDARDIZATION METHOD FOR THE  ANALYSIS
 OF GASEOUS ORGANIC COMPOUNDS
 Thomas  Bernstiel
 BCM Eastern
 A1r Resources Group
 One Plymouth Meeting
 Plymouth Meeting, PA  19462
     Standardization for the measurement of gaseous organic compounds,
according to the U.S. Environmental Protection Agency's Reference Method
18, is accomplished by using calibration gas cylinders or by preparing
known concentrations from liquids in Tedlar bags.  An alternative method
exists for the accurate preparation of standard gas concentrations from
pure liquids.  The method involves the use of a rigid container that can
be constructed of glass, aluminum, or stainless steel.  To prepare a stan-
dard gas concentration, the vessel is evacuated.  While under vacuum, a
known mass of the organic(s) of interest is injected, through a septum,
into the vessel.  The vessel is then pressurized to the desired level and
after mixing, the standard is ready for use.

     The advantages of this method of standard preparation as compared to
the Tedlar bag method include:  the determination of very accurate gas
volumes from pressure measurements, very accurate mass values for
injected organics because all liquid enters the evacuated vessel  and the
rapid and accurate preparation of up to six component organic mixtures.

     This technique has been used, under field conditions, to analyze
audit gas concentrations on the order of Ippm/v.
                                   945
-------
                        INTRODUCTION

    Gas  chromatographic  calibration  for  the  analyses  of
volatile organic  compounds  (VOC)  is accomplished, according
to  Environmental  Protection  Agency  Reference  Method  18,
Measurement  of  Gaseous  Organic  Compound  Emissions  by  Gas
Chromatography,  one of  two ways;  by  purchasing  certified
cylinders  containing   the   compound  or  mixtures  of  the
compounds  of  interest,  or by  preparing  standards  in tedlar
bags.  An  alternative method of standard preparation exists
that provides  the quality of data associated with cylinders
and  dilution   devices   while  allowing  the   flexibility
provided by the Tedlar Bag Method.

           Review of Existing Recommended Methods

    Certified  standards  are  available  for  a  wide  variety of
compounds,  and,   in general,   these   gases  are  of  high
quality;   but   variations  and   discrepancies  can   occur.
Typically, these  standard gases are shipped with  a  plus or
minus  2  percent  analytical  certifications  and  definite
information concerning  compound  stability.   Standard gases
are  rapid   and   easy   to   use  for   the   calibration   of
chromatographic  systems,  but  are   not without  their  own
specific   limitations,    both   logistical    and   technical.
First,   lead  time  on  receiving certified standard  gases is
on  the  order  of  4  to  6 weeks  for single  components,  and
longer  for mixtures.   This  time frame  can be  problematic
for many  practical  use  situations.   Certified  gases  and  gas
mixtures  can  also  be  expensive, ranging  from $70  to  $350
with  very  low  concentrations  costing  even more.   Prices
also vary widely from   supplier  to supplier  for the  same
product,  raising  some  concern  as   to  the  reason  for  the
price variations.   Consistent  with  this  concern  is the fact
that when a  standard  gas   is  purchased,  it is  only  one
concentration  and  must  be accepted  as  true.   Also, in order
to  verify response  over a  range or  ranges,  accommodation
must be  made  for dilution, or  multiple cylinders  must be
purchased.   Practical   volumes  is   another  very  important
consideration,  as  may  compounds of  potential  interest like
toluene, xylenes,  and  styrene are  vapor  restricted.  Vapor
restriction,  in this context,  refers   to  that property of
certain  volatile organic compounds  that relates  the vapor
pressure  of  the  compound at  some  selected temperature to
the  maximum   concentration  and   cylinder   pressure,  and,
therefore, gas volume,  that  can  be produced  without being
concerned   about  condensation.    The   equation   used   to
determine the degree of vapor restriction is  as follows:
                             946
-------
Equation 1

Vapor pressure at selected temperature (F) = Cylinder top
Target concentration (as percentage)off pressure
                                                (psi)

    Using  this  equation  predicts  the  feasibility  of  a
standard preparation  and  the volume that will  be produced,
which,  in   turn,   helps   determine  whether   the   use  of
cylinders  is  practical.   Specialty  gas vendors  also  use
this  equation  at  a  specified  temperature  in  order  to
determine the upper concentration limit that they certify.

    The  use of  tedlar bags to  prepare VOC  standards  has
certain advantages  not found  in the use of  gas  cylinders.
The  ability  to  prepare  standard  concentrations near  the
levels observed at  a  specific  source in field applications,
and  the  ability  to  prepare gaseous mixtures  in order  to
evaluate   system   and  column   performance   are   definite
advantages.  The major drawback  to  the  use  of bag standards
is   the   ability   to  produce   stable  and.  reproducible
concentrations.   Experience  has  shown,  and  bag  stability
data  will   support,   that  standards   produced   in  bags,
although suitable for  many applications, are  dependent upon
time,  bag   stability,  and the  specific  technique   used  to
fill the bags.   Even when the  best techniques are  used  to
fill bags,  some compounds,  like  styrene  and xylene,  present
vapor-pressure problems that can compromise  accuracy.

                  The  Alternative Method

    Having some benefits  of  both standardization  techniques
already mentioned  is  the  Rigid  Container  Technique.   When
using  the  Rigid  Container  Technique  for  calibration,  the
analyst is  able to  prepare  standards at the concentration
levels  actually  being observed.    In   the   case  of  mixed
solvent  systems,  specific   calibration  mixtures  of up  to
five or  six components can  be prepared  with concentration
accuracies  comparable to standard gas cylinders.

    The  vessel  of  choice  for   standard preparation  is  a
spherical  glass flask  of  approximately  6 liters  in  volume,
with  a  Teflon or  ground   glass  stopcock.  The flask  must
facilitate  the  injection  of  liquids  and  should  contain
three to  four small  glass  balls to  assist in gas  mixing.
Flask volumes  are  determined gravimetrically,  being  filled
with water and  weighed  several  times over a  period  of  days
using the  average  value  as  the  volume.   For  safety,  the
flask is wrapped with  good quality  tape, or  is dipped  in  a
plasticizer in case  the flask is  broken.
                             947
-------
    To  begin  the  preparation  of  a  standard,  the  glass
vessel  is  evacuated to  at  least 25  inches  of  Hg.   A known
mass  of  the  compound(s)  of   interest  is  then  injected
through  the  rubber septum with  a  microliter syringe.  When
this  is  done,  the entire mass of  the liquid is pulled into
the  flask  and  evaporated.   Next,  the  flask that  is still
under   a  hard   vacuum   and   now   contains  the  volatile
component(s)  is  connected  to  the  pressurization  system.
The   pressurization   system  is   composed   of   five  basic
elements,  including  a  balance  gas,  a  regulator,  a  U-tube
mercury  manometer,  a three-way  valve,  and  a  fine  metering
diaphram valve  with a  maximum delivery of  about 30 pounds
per   square   inch  (psi)  for   balance   gas  control.   The
delivery pressure of the system is  set using  the three-way
valve  and  the  mercury  U-tube  manometer  by orienting  the
three-way  valve  such that  the  flow  of balance  gas  to  the
head  of  the  manometer  and the desired  pressure  is  set with
the diaphram  valve.   Once the desired  delivery pressure is
set,  the three-way  valve is  changed to the position  that
will  direct the balance  gas to the flask,  filling it to the
value previously  set on  the mercury U-tube manometer.   The
time  required  for  the   flask  to  equilibrate   to  the  set
pressure is about 3 to 5 minutes.

    Filling the flask  produces an audible  sound.   When  the
pressure inside the flask reaches  atmospheric  pressure,  the
sound ceases.   At this  point, the flask should  be allowed
to  equilibrate  for  about 2 more  minutes.   After  allowing
sufficient time,  the  three-way  valve  is  oriented  so  that
the  now  pressurized  flask  is  connected  to   the  U-tube
manometer, and this  value is  recorded.   Ideally, the final
flask pressure should  be the  same as the set  pressure,  but
in  practice,  there  are  usually slight differences  on  the
order of 1 to 2 tenths  of a centimeter  of  mercury.   If  the
pressure of the flask has not  yet  come  to  the  set pressure,
simply switch the three-way valve  back  to the fill  position
and  wait  a   few  more  minutes,    or   calculate  the   gas
concentration based  on  the value  indicated.   The  equation
required for standard preparation by this method is:

Equation 2 - Concentration Calculation

ppm = ML  x  F  x  62363.6  x  K  x  106
             R  x  Cx  D

ML = Microliters injected (ul)
F = Density of liquid at 20 degrees C (g/ml)
K = Temperature of laboratory
H = Molecular weight of  liquid injected  (g)
C = Volume of flask (ml)
D = Total pressure of flask (mmHg)
62363.3 = Molar volume of any gas:  (22414 ml)(760 mm)
                                  273.15 (Std.  Temp. K)
                            948
-------
              Background Data and  Information

     In  order  to demonstrate how the technique  can  be used,
data  from  a typical field evaluation  is  presented  in Table
I.   The evaluation, which  produced  the  data  summarized in
Table  I,  was conducted  in the  summer of  1984 at  a semi-
conductor manufacturing facility.  The testing was observed
by  representatives  of the  state regulatory  agency,  who
provided  a  total  of  four  audit  samples.   These  audit
samples   were   prepared    onsite   using   a   high-quality
permiation  tube  system.    The   audits were  submitted  for
analysis in  tedlar bags  as two  different concentrations of
n-butyl   acetate   and  two   different   concentrations   of
acetone.   Table  II summarizes  the   results   of  the  audit
sample  analysis.   This  table  demonstrates  the  degree  of
accuracy   possible  using   this  method   of   calibration.
Comparison  of   the  percentage   differences   between  the
determined  values   and   true   values  yields   an  average
accuracy of 5 percent for the gas  chromatographic analysis.

    Table  I shows  the  preparation  of  multiple standards for
the  evaluation  and contains the inputs for Equation  2,  the
calculated  concentrations,  analytical  data (are count  and
instrument  attenuation),  and  normalized  response  factors.
The  normalized   response  factor  is  expressed  in  terms  of
parts  per  million  per  area unit (ppm/AU)  at a  specific
instrument  attenuation setting.  The attenuation  setting
selected is usually in the middle of  the range of settings
being  used.   Normalizing  the   standard  response  in  this
manner  uses  the  responses  to  observe  and  measure  the
standard   to    standard   variability.     Table   III    is   a
compound-specific  summary of  Table  I normalized  response
factors that  includes  a measure of percent variations from
the  average.  Agreement  between the average  response value
and  the  response produced by any  one  standard  is  typically
on the order of 5 percent, as Table III demonstrates.

                  Summary and Conclusions

    The  Rigid   Container  Technique for  the calibration  of
gas   chromatographic   systems   is  capable  of   producing
analytical  precision and accuracy  on the  order of  5  percent
under  field  conditions.    The   technique  facilitates  the
rapid  and  accurate  onsite preparation  of  VOC  calibration
standards  and   mixtures.   This   method of  calibration  gas
preparation utilizes variables  that  can  themselves be very
accurately  measured or defined.   For  example,  the  use  of
pressure  measurements   in  millimeters  of  mercury   and
gravimetrically  calibrated  containers  (to  0.1  ml)  adjusted
                            949
-------
to ambient  temperature  produces a  volume  value  superior to
that which  would be  produced  by a rotometer  and  stopwatch
or dry  gas  meter.   Also, the entire mass  of  compounds  that
has  been introduced  into  the  flask  is  volatized by  the
vacuum  in the  flask.  Another  advantage  of the technique is
that   since   standards   can   be  prepared   quickly   and
accurately,   calibration  standards   and   mixtures  can  be
prepared at  the same levels  observed in  the  sources  being
evaluated.   Further standard mixtures  can  contain  the  exact
compounds that  are  present  in  the  source samples,  which
allows   calibration  for   the   entire    mixture   in   one
injection.      Quantification   is   then    based   on    two
chromatrograms that are  essentially the same,  with respect
to  the  degree  of  tailing,   peak   overlap,  and   other
integration   parameters.   This  similarity   in  concentration
on and  composition  between  samples  and standards raises the
confidence level of the  data generated.

    In  conclusion,  the  inherent  accuracy  of  a  properly
constructed   pressurization  system,  the  type   of  analytical
precision and  accuracy   that  can be  produced  and the  low
cost to assemble a working  system the  rigid container  tech-
nique must be  considered a  viable calibration  technique  for
analytical   determinations   made  in  conjunction  with   EPA
Reference Method 18.
                            950
-------
                                                                                      TABLE  I
                                                                                FIELD DATA SUMMARY
CD
01
Factor
Liquid
Inspected Density
STD # Dt of Preps 1 Cmpnds Compounds Vol. (ul
1
2
3

4


5

6
7

7/30/84
7/30/84
7/30/84

7/30/84


7/31/84

7/31/84
7/31/84

2 Isopropanol
Freon
2 Isopropanol
Freon
2 Isopropanol
Freon
2 Isopropanol
Acetone
N-butyl
2 acetate
N-butyl
1 acetate
2 Acetone
Freon
0.9
0.8
0.85
0.85
2.40
2.30
1.15
1.05

1.20

0.80
1.10
1.00
) (ml)
0.785
1.560
0.785
1.560
0.785
1.560
0.785
0.791

0.882

0.882
0.791
1.560
Lab
Total
Molecular
Weight Volume
Temp(K) (g)
293
293
293
293
293
293
293
293

293

293
293
293
60.10
187.38
60.10
187.39
60.10
187.38
60.10
58.08

116.2

116.2
58.08
187.38
(ml)
6450.37
6450.37
6516.87
6516.87
6516.87
6516.87
6451.37
6450.37

6450.37

6516.87
6516.87
6516.87
Flask
Pressure
(mmHg)
1274.4
1274.4
1276.4
1276.4
1269.4
1269.4
1244.6
1244.6

1269.13

1281.37
1241.84
1241.84
Normal izecf
Flask Response
Concentration
(ppm)
26.13
14.80
24.46
15.60
69.25
42.30
34.23
32.62

20.33

13.30
33.83
18.75
PPM/AU
1
0
1
0
1
0
1
0

0

0
0
0
.100
.4716
.121
.4436
.232
.516
.550
.281

.318

.343
.265
.483
-------
        TABLE  II
   AUDIT DATA  SUMMARY
Acetone

True Value
(ppm)
1.46
39.10
Determined
Value
(ppm)
1.39
38.05
Percent
Difference
a>
1
3
n-Butvl

True Va
(ppm)
1.37
34.90
Acetate
Determined
lue Value
(ppm)
1.26
34,94

Percent
Difference
(%)
8
1
___.
       TABLE III
RESPONSE FACTOR SUMMARY

Freon
Normal Ized
Standard Response
Number Factor
1
2
3
8
Average

Standard
Number
5
8
Average
0.462
0.444
0.516
0.483
0.476
Acetone
Normal Ized
Response
Factor
0.281
0.265
0.273
Isopropanol
Difference
From Standard
Average Number
3 1
7 2
8 3
1 Averaae
5
n-Butvl Acetate
Percent
Difference
From Standard
Average Number
3 6
3 7
Average

Normal Ized
Response
Factor
1.100
1.121
1.232
1.151

Normal 1zed
Response
Factor
0.318
0.348
0.333

Percent
Difference
From
Average
4
3
7
5

Percent
Difference
From
Average
5
4
5
          952
-------
INTERPRETATION OF FIELD PERFORMANCE AUDIT DATA
IN WOODSTOVE EMISSION MEASUREMENT PROGRAMS
Joseph D. Evans, William M. Yeager, and
  Shrlkant V. Kulkarni
Center for Environmental Quality Assurance
Research Triangle Institute
Research Triangle Park, North Carolina  27709
Judith S. Ford and Robert C. McCr1ll1s
Air and Energy Engineering Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina  27711
    Over two heating seasons, the U.S. Environmental Protection Agency
(EPA), the New York State Energy Research and Development Authority
(NYSERDA), and the Coalition of Northeast Governors (CONEG) conducted
a study of the efficiency of four woodstove technologies: stoves
equipped with integral catalytic combustors, low-emission non-
catalytic stoves, retrofit catalyst woodstoves, and conventional
woodstoves.  A total of 68 homes were examined in the Glen Falls, New
York, and Waterbury, Vermont, areas.  Three principal  performance
parameters were examined in the study:

     • Emissions characteristics,
     • Wood-use characteristics, and

     • Creosote formation in flues.

     Field data were collected using an automated woodstove emission
sampler (AWES), and a data logger was connected to the AWES in order
to record various parameters.  A technical systems audit and a field
performance evaluation audit were conducted at the end of a testing
week during the second heating season.  The purpose of these audits
was to assess the Implementation of planned quality control activities
and the performance of measurement systems.  An oxygen monitoring
system (OMS), which operated with a micro-fuel cell, was one of the
critical measurement systems assessed.  Three separate oxygen check
concentrations were used to evaluate the OMS.  Two problems were
observed:  a drift in calibration and a negative bias at zero %
oxygen.

     To further evaluate the data and interpret the findings,
additional audits were conducted on subsequent AWES experiments
performed under controlled conditions in the laboratory.  It was found
that the negative bias observed in the field was data-logger-specific.
The drift, however,  appeared minimal under laboratory  conditions.
Where the drift was minimal,  data could be used after correcting for
the negative bias.   Findings of the audits demonstrated the Importance
of conducting performance evaluations with multiple check samples and
scheduling such audits appropriately in order that problems such as
drift may be detected and corrective actions Implemented.

                                   953
-------
Introduction

     Over two heating seasons, each lasting from about November
through March, a cooperative woodstove emission study was conducted by
the U.S. Environmental Protection Agency (EPA), the New York State
Energy Research and Development Authority (NYSERDA), and the Coalition
of Northeast Governors (CONEG).  The study was performed 1n 68
selected homes from the areas of Glen Falls, New York, and Waterbury,
Vermont.  Each home was equipped with a woodstove supplied by
different manufacturers.  Four types of woodstoves were compared
during the study:  1) stoves equipped with integral catalytic
combustors, 2) low-emission non-catalytic stoves, 3) stoves with a
retrofit catalyst, and 4) conventional woodstoves.

     Three principal stove performance characteristics were examined
in the study:  emissions (particulates and selected polynuclear
aromatic hydrocarbons), wood-use characteristics, and creosote
formation 1n the woodstove flue.  Based upon these criteria, a major
objective of the study was to rank stove types according to
performance.

     In order to assess data quality, a field audit was conducted by
EPA's A1r and Energy Engineering Research Laboratory (AEERL) quality
assurance (QA) program personnel.  This audit included a performance
evaluation of critical measurements using reference materials, and a
technical systems evaluation of equipment,  standard operating
procedures, personnel capabilities, and adherence to the approved
AEERL quality assurance project plan (QAPP).


Experimental Design

     To sample the flue gases, an automated woodstove emission sampler
(AWES)  was set up to run in each home for 7 days, sampling for 1-
minute periods at 30-mlnute Intervals.  The AWES unit operated on a
timed cycle with a pump drawing stack samples from a probe 1n the
center of the woodstove flue, then through  a filter, an XAD-2
cartridge, and finally an oxygen sensing device.  Flow was regulated,
using a calibrated critical  orifice,  at approximately 1 liter of flue
gas per minute.

     Total partlculate collection consisted of partlculate from the
filter as well as from rinses of sample lines, filter cartridge, and
sample probe.

     The oxygen sensing device was a simple micro-fuel cell.  Near the
end of the 1-minute sampling period,  after sufficient flue gas had
flushed the cell, an oxygen reading was taken and recorded by a data
logger attached to the AWES.

     Other functions of the data logger Included periodic recordings
of flue gas temperature, stove temperature,  and room temperature from
thermocouples connected to the unit.   Because the data logger also
recorded total wood use, the homeowner was  required to place all wood
on a calibrated scale connected to the data logger before putting the
wood Into the stove.  Wood-loading patterns by the homeowner could
then be correlated with other collected Information such as stack
temperature and stack oxygen concentration.

                                   954
-------
Calibration

     Calibration and quality control  (QC) procedures were conducted at
the beginning of each testing period.  Two technicians would Install
an AWES 1n a selected home for 1 week.  Installation procedures
Included calibration of the wood scale, a QC check of the flow through
the critical orifice, and calibration of the oxygen micro-fuel cell.
Weights were calibrated using a series of selected weights placed on
the wood scale and adjusting the calibration curve stored 1n the data
logger accordingly.  The flow was checked with a rotameter, and this
QC value was recorded.  A limit for the range of acceptable flow
values was established 1n the QAPP.

     The oxygen sensor was calibrated with a single concentration
value of 20.9%, the standard oxygen concentration 1n ambient air.
This made oxygen calibration very simple since it required no
certified gas cylinders.  Theoretically, an oxygen fuel cell has a
linear response.  If there is no oxygen 1n the gas surrounding the
cell, no oxidation occurs and no current Is produced.  Therefore, the
fuel cell has an absolute zero value and should require only a one-
point calibration.

     For this study, a data logger recorded the cell voltage (V) for
ambient air and set this value Internally at 20.9% oxygen.  Voltages
produced during sampling were converted to 02 concentrations by a
simple scaling formula:

     %02 - 20.9% (Vx/V2o.g).

Because the oxidation reaction 1s temperature-dependent, these fuel
cells had temperature compensation circuitry built Into the cell.

     At the end of the testing week, calibration was checked for the
wood scale using a series of weights,  for the critical  orifice using a
rotameter,  and for the oxygen micro-fuel cell  using ambient air.
Results of these checks were recorded.  QC results from the end of the
testing week were checked to ensure they were within specifications
given 1n the QAPP.


Field Audits

     Along  with a technical  systems audit,  a performance evaluation
audit was performed as an Integral  part of the woodstove program.
Four measurement systems were chosen for the audit:

     • Wood-scale calibration was audited using four NBS-traceable
       weights.

     • The  gas flow through the critical  orifice was audited with a
       bubble flowmeter.   This was  checked three times.

     • Oxygen concentrations recorded  by the data logger were audited
       using two certified gas cylinders.   Three concentrations were
       used Including a zero gas  (nitrogen),  a 10.2% oxygen  cylinder,
       and  ambient  air at 20.9% since  1t was readily available.

     • Thermocouple temperatures  recorded by the data logger were
       checked with a solid state temperature  probe.

                                   955
-------
     Audit parameters were chosen based on the critical nature of the
measurement.  An analysis of error propagation determined that flow
rate and oxygen concentration were the most critical.


Woodstove Emission  Values

     Standard units  for reporting woodstove emission data are:

     1. Mass of particulate/mass dry wood burned  (mg/kg wood)
     2. Mass of partlculate/heat output (mg/Btu)
     3. Mass of partlculate/unit time (mg/hour)
     4. Mass of particulate/volume of gas (mg/llter flue gas)

     The last of these 1s the easiest to measure; 1t can be calculated
directly from the gas sample which passes through the AWES sampler.

Mass of partlculate  =            mass of parti cul ate (MP)	^_^
  volume of gassampler flow rate (FR) x sampling duration (SD)


where:

     MP = total partlculate collected on the filter, 1n the sample
          probe, 1n  associated sample lines, etc.

     FR = sampling  rate which 1s calibrated for each Individual AWES
          (approximately 1 Hter/m1n)

     SD = total sample collection time (one l-m1nute sample every 30
          minutes for 7 days = 336 minutes)

     Calculation of  the partlculate emissions 1n any of the other
units requires knowledge of the total emissions from the stove over
the sampling period.  Since the AWES sampler 1s not 1sok1net1c, the
measured partlculate mass represents an unknown fraction of the total
flue gas volume (TFGV).  Therefore, calculation of the total  emissions
requires determination of the TFGV.  Then

total mass of partlculate emissions = mass of partlculate collected (MPJ
  total flue gas volume (TFGV)sample volume (FR • SD)

The simplest way to determine the TFGV 1s to Integrate the flue gas
flow rate over the sampling period.  The flue gas flow rate cannot be
measured directly,  however,  because 1t 1s so small for much of the
time.  In this study, the TFGV was calculated from the stolchlometrlc
volume (SV)  of the wood burned and the volume of the excess air (EA)
over what Is required for combustion.

     The SV is the volume of gas produced by complete combustion  of 1
kg of wood.   It includes the non-oxygen  component of the atmospheric
air required for complete combustion. Wood 1s composed  of carbon,
oxygen, hydrogen,  and nitrogen (C,  0, H,  N).   During combustion,  these
oxidize to CO,  C02,  H20,  and N?0.   The oxygen component  of wood 1s
insufficient to completely oxidize the other elements in the  wood.
Additional  oxygen  must come from the atmosphere.   Using  the known
elemental  composition of wood,  the volume of air required for complete
combustion can be  calculated.   The volume of N£(  Ar,  etc.  1s  added to

                                   956
-------
 the  volume  of CO,  C02,  H20,  and  N02  produced by  combustion to
 determine the stolchlometrlc volume.

      The TFGV Is  not  stmply  the  SV.   If  1t were, the  flue gas would
 contain no  02.   In practice,  more  air enters the stove than  1s
 required for combustion.   Oxygen 1s  20.9% of the atmosphere;
 therefore,  TFGV  can be  calculated  by:

      TFGV = SV  [20.9%/(20.9% - % 02  measured)]

 where % 02  measured equals the measured  02 1n the woodstove  flue.
 Then,

      total  emissions  =  MP  •  TFGV/(FR • SD).

      The following example Illustrates how a small error 1n  the 02
 measurement may  significantly affect the calculation  of total
 emissions.   Assume that the  average  true 02 value during a test period
 1s 15% and  that  the average  measured value 1s 17%.  This represents an
 absolute error of  only  2%  02 or  a  relative error of 13%.  Using the
 equation to calculate TFGV,  1f the 02  concentration 1s 15% (true), we
 have  TFGV = SV(20.9/5.9) = 3.54  SV.   If  the 02 concentration 1s 17%
 (measured),  then TFGV = SV (20.9/3.9)  =  5.36 SV.  Accuracy (containing
 both  random and  systematic error)  can  be defined for  a single
 measurement as

      measured cone. - true cone, x 100 = 5.36 - 3.54  SV x 100 = 51.4%
          true cone.                         3.54 SV

 This  much error  1n the  reported  emission values could have significant
 Impact on ranking  the stove  types  based  on emissions.


 Field Audit Results

      Results  of the oxygen performance evaluation audit follow:

 Audit cylinder and            Value measured
 ambient concentration        by  micro-fuel cell

        0.0%                       -2.2%
        10.2%                      10.4%
        20.9%                      23.8%
Note that, while the micro-fuel cell measurement 1s within 0.2% of the
midpoint audit value, the ambient concentration has drifted upscale
+2.9% and, even more significantly, the zero concentration measures
-2.2% oxygen.  Figure 1 compares the field audit results to the Ideal
response of the oxygen sensor.

    These results raised questions concerning the positive drift of
the oxygen value at ambient concentration and,  1n particular,
measurement of a negative concentration at zero % oxygen.   The theory
behind a micro-fuel cell suggests that negative values at  zero %
oxygen concentration would be Impossible unless there was  a systematic
negative bias In the measurement system.  This  was 1n conflict with
the positive drift occurring at ambient concentration.  If It  1s

                                   957
-------
assumed that the calibration at the beginning of the week was correct
then, sometime after this, there was a positive drift at the upper end
of the calibration curve.  It was decided that a performance
evaluation audit of AWES units In the laboratory would be undertaken
to more completely understand the results of the field audit.
Laboratory Audit Results

    Laboratory testing for woodstove emissions was performed similarly
to field testing with a Modified Method 5 train and ASTM dilution
tunnel added to the woodstove flue for comparison purposes.   Two
Intermittent AWES units (similar to field operations)  and one
continually operating AWES were in operation on one woodstove.   The
audit took place during the middle of the testing week;  therefore,  the
Intermittent-operating AWES units were calibrated 4 days prior  and  the
continuous AWES unit was replaced and calibrated regularly every 8
hours.

    Four calibration gases were used for the oxygen performance audit:
nitrogen containing zero % oxygen, a 9.7% oxygen cylinder, a 15.2%
oxygen cylinder, and ambient concentration at 20.9% oxygen.   Results
of these four concentrations for each AWES are presented In  Tables  I,
II, and III.  As in the field audit, all units showed  a  negative
reading at zero % oxygen.  There was a noticeable error  for all units
at the ambient oxygen concentration:  for two units, the error  was
positive; for the other, the error was negative.  Although the  oxygen
drift was much less than the AWES unit audited in the  field and
presented less concern for the quality of data, the negative reading
at zero % oxygen for all units audited was still disturbing.
Data Audit

    An audit of data quality was also performed for this project.  The
purpose of this audit was to assess the methods used to collect,
interpret, and report the Information required to characterize data
quality.


    Analysis of these three types of audits showed that:

     • A negative voltage bias was produced by each AWES data logger.
       This bias was different for each unit; consequently, oxygen
       readings were 1n error.

     • The oxygen micro-fuel cell calibration drifted during the
       testing week.  In the AWES field units, this drift at ambient
       oxygen concentration was both large and small, while laboratory
       units showed minimal drift.  The drift was both positive and
       negative; while the exact reasons for this drift remain
       unknown, these variations are significant when computing
       woodstove emission data.
                                   958
-------
Conclusions

     Important conclusions were formulated concerning auditing
procedures and ways 1n which auditors can aid researchers 1n Improving
data quality.  These conclusions follow:

     1. A multipoint performance evaluation audit proved to be
        Important for what was known to be a linear measurement.  For
        example:

        a. If only a midpoint check (10% 02) had been performed, the
           problem with the oxygen sensor would not have been
           detected.

        b. If only two points had been checked [e.g., 10% 0? and
           ambient (20.9% 0?)], the seriousness of the problem would
           have remained unknown; 1n particular,  the negative bias of
           the data logger.  It would have probably been Incorrectly
           assumed that the 03 measurement was non-linear at higher
           oxygen concentrations.

As a result of a three-point audit (zero, 10%, and ambient 03) and a
detailed analysis of audit results,  the negative  bias of the data
logger was discovered,  and a random drift was Identified 1n the oxygen
micro-fuel cells.

     2. The audit results prompted the contractor to find out the
        reason for the negative voltage bias.  Where the oxygen drift
        was minimal,  data could be corrected after adjusting for this
        bias caused by Individual data loggers.

     3. Performance evaluation audits should be performed well after
        calibration of an Instrument 1n order to  account for
        Instrument drift that may occur with time or changing
        conditions.

     4. A performance evaluation audit can Identify measurement system
        error.  A technical systems  audit may be  required,  however,  to
        find the source of that error.  An audit  of data quality can
        be performed to determine the effect of the error on the
        quality of the data reported.
                                  959
-------
  TABLE I.  RESULTS OF FLUE GAS OXYGEN AUDIT, AWES
             INTERMITTENT SAMPLER NO.  23
Audit Sample
% 02
0.0
9.7
15.2
20.9
Data Logger Reading
% 02
- 0.8
9.4
15.2
21.8
D1 f f erence
% 02
- 0.8
- 0.3
0.0
0.9
  TABLE II.  RESULTS OF FLUE GAS OXYGEN AUDIT, AWES
             INTERMITTENT SAMPLER NO.  27
Audit Sample
% 02
0.0
9.7
15.2
20.9
Data Logger Reading
% 02
- 0.8
9.3
15.0
21.4
Difference
% 02
- 0.8
- 0.4
- 0.2
0.5
 TABLE III.  RESULTS OF FLUE GAS  OXYGEN AUDIT, AWES
              CONTINUOUS SAMPLER  NO.  13
Audit Sample       Data Logger Reading        Difference
    % 02                  %  02                   % 02

     O                 ^T7l                 ^T7l
     9.7                  8.7                 - 1.0
    15.2                  14.3                 - 0.9
    20.9                  20.6                 - 0.3
                          960
-------
c
o
o>
X
X
o
T>
0

3
n
o
9
        -2
        -4
                                                         o audit results


                                                         4-  ideal response
                    r~  i    i    ii	r—i	r~—i	1	1	1    i

   2       4        6        8      10       12      14      16      18


                            Known % Oxygen

Figure 1.   Audit results versus  ideal oxygen sensor response.
                                                                                              20
-------
                          INDEX
Accidental Releases
   733
Acidic Deposition
   164,170,176,182,189,200,216,
   227,237,243
Activity Questionnaire
   720
Adsorbent Tubes
   670
Aerosols
   699
Aldehydes
   814
Ambient Measurement
   1,251,259,265,277 ,285,291,299 ,
   305,313,324,331,341,352,358,369
Ammonia
   170
Analysis
   305
Analytical Criteria
   613
Asbestos
   645
Atmospheric Carbon
   853
Audits
   953
Automation
   265
B

Benzene
   814
Cloud Water Chemistry
   237
Cogener Profiles
   634
Combustion Sources
   634,769
Combustion Spillage
   727
Comparison
   324
Control Strategy
   1
Cotinine
   155
Cryogenic Trapping
   265,285, 305
D

Data Validation
   922
Denuder Systems
   170,685,691
Detection Limits
   613,922
Diesel Exhaust
   119
Dilution Bottles
   285
Dioxins
   590,634
Dry Deposition
   170,189,200,699
Dust
   922
Carbon Molecular Sieves
   670
Carbon-Based Adsorbents
   670
Carcinogens
   922
Certification
   945
Chemical Artifacts
   699
Chemometrics
   548,556,569,575
                                             Ethylene Oxide
                                                524,530
Forests
   189,237
Formaldehyde
   814
Furans
   590
                                    963
-------
Gas Chromatography
   15,155,265,305,945
Gas Cylinders
   265,945
H
Hazardous Waste
   383,399,406,413,418,426,432,441
   447,461,470,486,670
Hospitals
   524
Hydrocarbons
   814
Incinerators
   634
Indoor Air Quality
   84,89,98,104,113,119,131,137,
   143,149,155,715,814
Integrated Air Cancer Project
   15,799,804,814,821,828,835,841,
   864,870,879,885,890
Integrated Environmental
   Management
   377
                                             Mutagenicity
                                                853,879
N
National Air Toxics Program
   1
Netherlands
   691
Network Design
   341
Nicotine
   155
Nitrate
   170
Nitric Acid
   170
Nitrogenous Particles
   699
Nitropyrene
   119
Organics
   4,7,15,21,27,34,42,51,57,63,72,
   78,769
K

Keynote
   1
Love Canal
   896
M
Method Evaluation
   775
Methyl Chloroform
   324
Multiple Linear Regression
   853
Municipal Waste
   634
Particulates
   953
Peroxyacetyl Nitrate
   679
Polychlorinated Biphenyls
   922
Polycyclic Aromatic
   Hydrocarbons
   151
Polyurethane Foam
   590
Pyrolysis
   670
Quality Assurance
   613,634,896,914,922,932,939,
   945,953
                                    964
-------
Residential Sampling
   879
Risk Assessment
   922
U

Uncertainty
   265
Sample Analysis
   655
Sampling Efficiency
   590
Scrubber Efficiency
   530
Semivolatiles
   4,7,15,21,27,34,42,51,57,63,72,
   78,670
Simulated Atmospheres
   685
Source Apportionment
   853
Source Measurement
   1
Source Monitoring
   497,503,510,517,524,530,536,541
Spectrometry
   15,285,750,775,896
Stack Gas Emissions
   613
Statistical Analysis
   556,575
Sterilization  Chambers
   524,530
Sulfate
   170
Sulfur Dioxide
   640
Superfund
   922
Volatile Organics
   119,739,750,765,769,775,787,793
   670,945
Volatilization
   922
W
Water Vapor
   285
Wet Deposition
   237
Woodburning
   15,664,814,853,879,953
Workplace
   119
 TCDD/TCDF
    590,596,602,613,621,629,634
 Thermal Desorption
    285
 Tobacco Smoke
    155
 Toluene
    324
                                      965
-------